1 /* $NetBSD: ip_icmp.c,v 1.179 2025/02/22 09:10:27 mlelstv Exp $ */
2
3 /*
4 * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Public Access Networks Corporation ("Panix"). It was developed under
9 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
10 *
11 * This code is derived from software contributed to The NetBSD Foundation
12 * by Jason R. Thorpe of Zembu Labs, Inc.
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
24 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
25 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
27 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33 * POSSIBILITY OF SUCH DAMAGE.
34 */
35
36 /*
37 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
38 * All rights reserved.
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 * 3. Neither the name of the project nor the names of its contributors
49 * may be used to endorse or promote products derived from this software
50 * without specific prior written permission.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * SUCH DAMAGE.
63 */
64
65 /*
66 * Copyright (c) 1982, 1986, 1988, 1993
67 * The Regents of the University of California. All rights reserved.
68 *
69 * Redistribution and use in source and binary forms, with or without
70 * modification, are permitted provided that the following conditions
71 * are met:
72 * 1. Redistributions of source code must retain the above copyright
73 * notice, this list of conditions and the following disclaimer.
74 * 2. Redistributions in binary form must reproduce the above copyright
75 * notice, this list of conditions and the following disclaimer in the
76 * documentation and/or other materials provided with the distribution.
77 * 3. Neither the name of the University nor the names of its contributors
78 * may be used to endorse or promote products derived from this software
79 * without specific prior written permission.
80 *
81 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
82 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
83 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
84 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
85 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
86 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
87 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
88 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
89 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
90 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
91 * SUCH DAMAGE.
92 *
93 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
94 */
95
96 #include <sys/cdefs.h>
97 __KERNEL_RCSID(0, "$NetBSD: ip_icmp.c,v 1.179 2025/02/22 09:10:27 mlelstv Exp $");
98
99 #ifdef _KERNEL_OPT
100 #include "opt_ipsec.h"
101 #endif
102
103 #include <sys/param.h>
104 #include <sys/systm.h>
105 #include <sys/mbuf.h>
106 #include <sys/protosw.h>
107 #include <sys/socket.h>
108 #include <sys/socketvar.h> /* For softnet_lock */
109 #include <sys/kmem.h>
110 #include <sys/time.h>
111 #include <sys/kernel.h>
112 #include <sys/syslog.h>
113 #include <sys/sysctl.h>
114
115 #include <net/if.h>
116 #include <net/route.h>
117
118 #include <netinet/in.h>
119 #include <netinet/in_systm.h>
120 #include <netinet/in_var.h>
121 #include <netinet/ip.h>
122 #include <netinet/ip_icmp.h>
123 #include <netinet/ip_var.h>
124 #include <netinet/in_pcb.h>
125 #include <netinet/in_proto.h>
126 #include <netinet/icmp_var.h>
127 #include <netinet/icmp_private.h>
128 #include <netinet/wqinput.h>
129
130 #ifdef IPSEC
131 #include <netipsec/ipsec.h>
132 #include <netipsec/key.h>
133 #endif
134
135 /*
136 * ICMP routines: error generation, receive packet processing, and
137 * routines to turnaround packets back to the originator, and
138 * host table maintenance routines.
139 */
140
141 int icmpmaskrepl = 0;
142 int icmpbmcastecho = 0;
143 int icmpreturndatabytes = 8;
144
145 percpu_t *icmpstat_percpu;
146
147 /*
148 * List of callbacks to notify when Path MTU changes are made.
149 */
150 struct icmp_mtudisc_callback {
151 LIST_ENTRY(icmp_mtudisc_callback) mc_list;
152 void (*mc_func)(struct in_addr);
153 };
154
155 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks =
156 LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks);
157
158 /* unused... */
159 u_int ip_next_mtu(u_int, int);
160
161 bool icmp_dynamic_rt_msg = false;
162
163 static int icmperrppslim = 100; /* 100pps */
164 static int icmperrpps_count = 0;
165 static struct timeval icmperrppslim_last;
166 static int icmp_rediraccept = 1;
167 static int icmp_redirtimeout = 600;
168 static struct rttimer_queue *icmp_redirect_timeout_q = NULL;
169
170 /* Protect mtudisc and redirect stuff */
171 static kmutex_t icmp_mtx __cacheline_aligned;
172
173 static void icmp_send(struct mbuf *, struct mbuf *);
174 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *);
175 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *);
176
177 static void sysctl_netinet_icmp_setup(struct sysctllog **);
178
179 /* workqueue-based pr_input */
180 static struct wqinput *icmp_wqinput;
181 static void _icmp_input(struct mbuf *, int, int);
182
183 void
icmp_init(void)184 icmp_init(void)
185 {
186
187 sysctl_netinet_icmp_setup(NULL);
188
189 mutex_init(&icmp_mtx, MUTEX_DEFAULT, IPL_NONE);
190 /*
191 * This is only useful if the user initializes redirtimeout to
192 * something other than zero.
193 */
194 mutex_enter(&icmp_mtx);
195 icmp_redirect_timeout_q = rt_timer_queue_create(icmp_redirtimeout);
196 mutex_exit(&icmp_mtx);
197
198 icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS);
199 icmp_wqinput = wqinput_create("icmp", _icmp_input);
200 }
201
202 void
icmp_mtudisc_lock(void)203 icmp_mtudisc_lock(void)
204 {
205
206 mutex_enter(&icmp_mtx);
207 }
208
209 void
icmp_mtudisc_unlock(void)210 icmp_mtudisc_unlock(void)
211 {
212
213 mutex_exit(&icmp_mtx);
214 }
215
216 /*
217 * Register a Path MTU Discovery callback.
218 */
219 void
icmp_mtudisc_callback_register(void (* func)(struct in_addr))220 icmp_mtudisc_callback_register(void (*func)(struct in_addr))
221 {
222 struct icmp_mtudisc_callback *mc, *new;
223
224 new = kmem_alloc(sizeof(*mc), KM_SLEEP);
225
226 mutex_enter(&icmp_mtx);
227 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
228 mc = LIST_NEXT(mc, mc_list)) {
229 if (mc->mc_func == func) {
230 mutex_exit(&icmp_mtx);
231 kmem_free(new, sizeof(*mc));
232 return;
233 }
234 }
235
236 new->mc_func = func;
237 LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, new, mc_list);
238 mutex_exit(&icmp_mtx);
239 }
240
241 /*
242 * Generate an error packet of type error in response to a bad IP packet. 'n'
243 * contains this packet. We create 'm' and send it.
244 *
245 * As we are not required to return everything we have, we return whatever
246 * we can return at ease.
247 *
248 * Note that ICMP datagrams longer than 576 octets are out of spec according
249 * to RFC1812; the limit on icmpreturndatabytes will keep things below that
250 * limit.
251 */
252 void
icmp_error(struct mbuf * n,int type,int code,n_long dest,int destmtu)253 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
254 {
255 struct ip *oip = mtod(n, struct ip *), *nip;
256 const unsigned oiphlen = oip->ip_hl << 2;
257 struct icmp *icp;
258 struct mbuf *m;
259 struct m_tag *mtag;
260 unsigned datalen, mblen;
261 int totlen;
262
263 if (type != ICMP_REDIRECT)
264 ICMP_STATINC(ICMP_STAT_ERROR);
265
266 /*
267 * Don't send error if:
268 * - The original packet was encrypted.
269 * - The packet is multicast or broadcast.
270 * - The packet is not the first fragment of the message.
271 * - The packet is an ICMP message with an unknown type.
272 */
273 if (n->m_flags & M_DECRYPTED)
274 goto freeit;
275 if (n->m_flags & (M_BCAST|M_MCAST))
276 goto freeit;
277 if (oip->ip_off &~ htons(IP_MF|IP_DF))
278 goto freeit;
279 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
280 n->m_len >= oiphlen + ICMP_MINLEN) {
281 struct icmp *oicp = (struct icmp *)((char *)oip + oiphlen);
282 if (!ICMP_INFOTYPE(oicp->icmp_type)) {
283 ICMP_STATINC(ICMP_STAT_OLDICMP);
284 goto freeit;
285 }
286 }
287
288 /*
289 * First, do a rate limitation check.
290 */
291 if (icmp_ratelimit(&oip->ip_src, type, code)) {
292 /* XXX stat */
293 goto freeit;
294 }
295
296 /*
297 * Compute the number of bytes we will put in 'icmp_ip'. Truncate
298 * it to the size of the mbuf, if it's too big.
299 */
300 datalen = oiphlen + uimin(icmpreturndatabytes,
301 ntohs(oip->ip_len) - oiphlen);
302 mblen = 0;
303 for (m = n; m && (mblen < datalen); m = m->m_next)
304 mblen += m->m_len;
305 datalen = uimin(mblen, datalen);
306
307 /*
308 * Compute the total length of the new packet. Truncate it if it's
309 * bigger than the size of a cluster.
310 */
311 CTASSERT(ICMP_MINLEN + sizeof(struct ip) <= MCLBYTES);
312 totlen = sizeof(struct ip) + ICMP_MINLEN + datalen;
313 if (totlen > MCLBYTES) {
314 datalen = MCLBYTES - ICMP_MINLEN - sizeof(struct ip);
315 totlen = MCLBYTES;
316 }
317
318 /*
319 * Allocate the mbuf for the new packet.
320 */
321 MGETHDR(m, M_DONTWAIT, MT_HEADER);
322 if (m && (totlen > MHLEN)) {
323 MCLGET(m, M_DONTWAIT);
324 if ((m->m_flags & M_EXT) == 0) {
325 m_freem(m);
326 m = NULL;
327 }
328 }
329 if (m == NULL)
330 goto freeit;
331 MCLAIM(m, n->m_owner);
332 m->m_len = totlen;
333 m->m_pkthdr.len = m->m_len;
334 m_copy_rcvif(m, n);
335
336 if ((u_int)type > ICMP_MAXTYPE)
337 panic("icmp_error");
338 ICMP_STATINC(ICMP_STAT_OUTHIST + type);
339
340 if ((m->m_flags & M_EXT) == 0)
341 m_align(m, m->m_len);
342
343 /*
344 * Get pointers on the IP header and the ICMP header.
345 */
346 nip = mtod(m, struct ip *);
347 icp = (struct icmp *)(nip + 1);
348
349 /*
350 * Fill in the fields of the ICMP header: icmp_type, icmp_code
351 * and icmp_ip. icmp_cksum gets filled later.
352 */
353 icp->icmp_type = type;
354 if (type == ICMP_REDIRECT) {
355 icp->icmp_gwaddr.s_addr = dest;
356 } else {
357 icp->icmp_void = 0;
358 /*
359 * The following assignments assume an overlay with the
360 * zeroed icmp_void field.
361 */
362 if (type == ICMP_PARAMPROB) {
363 icp->icmp_pptr = code;
364 code = 0;
365 } else if (type == ICMP_UNREACH &&
366 code == ICMP_UNREACH_NEEDFRAG && destmtu)
367 icp->icmp_nextmtu = htons(destmtu);
368 }
369 icp->icmp_code = code;
370 m_copydata(n, 0, datalen, (void *)&icp->icmp_ip);
371
372 /*
373 * Now, copy the old IP header (without options) in front of the
374 * ICMP message. The src/dst fields will be swapped in icmp_reflect.
375 */
376 /* ip_v set in ip_output */
377 nip->ip_hl = sizeof(struct ip) >> 2;
378 nip->ip_tos = 0;
379 nip->ip_len = htons(m->m_len);
380 /* ip_id set in ip_output */
381 nip->ip_off = htons(0);
382 /* ip_ttl set in icmp_reflect */
383 nip->ip_p = IPPROTO_ICMP;
384 nip->ip_src = oip->ip_src;
385 nip->ip_dst = oip->ip_dst;
386 /* move PF m_tag to new packet, if it exists */
387 mtag = m_tag_find(n, PACKET_TAG_PF);
388 if (mtag != NULL) {
389 m_tag_unlink(n, mtag);
390 m_tag_prepend(m, mtag);
391 }
392
393 icmp_reflect(m);
394
395 freeit:
396 m_freem(n);
397 }
398
399 struct sockaddr_in icmpsrc = {
400 .sin_len = sizeof(struct sockaddr_in),
401 .sin_family = AF_INET,
402 };
403
404 /*
405 * Process a received ICMP message.
406 */
407 static void
_icmp_input(struct mbuf * m,int hlen,int proto)408 _icmp_input(struct mbuf *m, int hlen, int proto)
409 {
410 struct icmp *icp;
411 struct ip *ip = mtod(m, struct ip *);
412 int icmplen;
413 int i;
414 struct in_ifaddr *ia;
415 void *(*ctlfunc)(int, const struct sockaddr *, void *);
416 int code;
417 struct rtentry *rt;
418 struct sockaddr_in icmpdst = {
419 .sin_len = sizeof(struct sockaddr_in),
420 .sin_family = AF_INET,
421 };
422 struct sockaddr_in icmpgw = {
423 .sin_len = sizeof(struct sockaddr_in),
424 .sin_family = AF_INET,
425 };
426
427 /*
428 * Locate icmp structure in mbuf, and check
429 * that not corrupted and of at least minimum length.
430 */
431 icmplen = ntohs(ip->ip_len) - hlen;
432 if (icmplen < ICMP_MINLEN) {
433 ICMP_STATINC(ICMP_STAT_TOOSHORT);
434 goto freeit;
435 }
436 i = hlen + uimin(icmplen, ICMP_ADVLENMIN);
437 if (M_UNWRITABLE(m, i) && (m = m_pullup(m, i)) == NULL) {
438 ICMP_STATINC(ICMP_STAT_TOOSHORT);
439 return;
440 }
441 ip = mtod(m, struct ip *);
442 m->m_len -= hlen;
443 m->m_data += hlen;
444 icp = mtod(m, struct icmp *);
445 /* Don't need to assert alignment, here. */
446 if (in_cksum(m, icmplen)) {
447 ICMP_STATINC(ICMP_STAT_CHECKSUM);
448 goto freeit;
449 }
450 m->m_len += hlen;
451 m->m_data -= hlen;
452
453 if (icp->icmp_type > ICMP_MAXTYPE)
454 goto raw;
455 ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type);
456 code = icp->icmp_code;
457
458 switch (icp->icmp_type) {
459 case ICMP_UNREACH:
460 switch (code) {
461 case ICMP_UNREACH_PROTOCOL:
462 code = PRC_UNREACH_PROTOCOL;
463 break;
464
465 case ICMP_UNREACH_PORT:
466 code = PRC_UNREACH_PORT;
467 break;
468
469 case ICMP_UNREACH_SRCFAIL:
470 code = PRC_UNREACH_SRCFAIL;
471 break;
472
473 case ICMP_UNREACH_NEEDFRAG:
474 code = PRC_MSGSIZE;
475 break;
476
477 case ICMP_UNREACH_NET:
478 case ICMP_UNREACH_NET_UNKNOWN:
479 case ICMP_UNREACH_NET_PROHIB:
480 case ICMP_UNREACH_TOSNET:
481 code = PRC_UNREACH_NET;
482 break;
483
484 case ICMP_UNREACH_HOST:
485 case ICMP_UNREACH_HOST_UNKNOWN:
486 case ICMP_UNREACH_ISOLATED:
487 case ICMP_UNREACH_HOST_PROHIB:
488 case ICMP_UNREACH_TOSHOST:
489 case ICMP_UNREACH_ADMIN_PROHIBIT:
490 case ICMP_UNREACH_HOST_PREC:
491 case ICMP_UNREACH_PREC_CUTOFF:
492 code = PRC_UNREACH_HOST;
493 break;
494
495 default:
496 goto badcode;
497 }
498 goto deliver;
499
500 case ICMP_TIMXCEED:
501 if (code > 1)
502 goto badcode;
503 code += PRC_TIMXCEED_INTRANS;
504 goto deliver;
505
506 case ICMP_PARAMPROB:
507 if (code > 1)
508 goto badcode;
509 code = PRC_PARAMPROB;
510 goto deliver;
511
512 case ICMP_SOURCEQUENCH:
513 if (code)
514 goto badcode;
515 code = PRC_QUENCH;
516 goto deliver;
517
518 deliver:
519 /*
520 * Problem with datagram; advise higher level routines.
521 */
522 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
523 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
524 ICMP_STATINC(ICMP_STAT_BADLEN);
525 goto freeit;
526 }
527 if (m->m_len < hlen + ICMP_ADVLEN(icp)) {
528 m = m_pullup(m, hlen + ICMP_ADVLEN(icp));
529 if (m == NULL)
530 goto freeit;
531 }
532 ip = mtod(m, struct ip *);
533 icp = (struct icmp *)(mtod(m, uint8_t *) + hlen);
534
535 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
536 goto badcode;
537
538 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
539 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
540 if (ctlfunc)
541 (void) (*ctlfunc)(code, sintosa(&icmpsrc),
542 &icp->icmp_ip);
543 break;
544
545 badcode:
546 ICMP_STATINC(ICMP_STAT_BADCODE);
547 break;
548
549 case ICMP_ECHO:
550 if (!icmpbmcastecho &&
551 (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
552 ICMP_STATINC(ICMP_STAT_BMCASTECHO);
553 break;
554 }
555 icp->icmp_type = ICMP_ECHOREPLY;
556 goto reflect;
557
558 case ICMP_TSTAMP:
559 if (icmplen < ICMP_TSLEN) {
560 ICMP_STATINC(ICMP_STAT_BADLEN);
561 break;
562 }
563 if (!icmpbmcastecho &&
564 (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
565 ICMP_STATINC(ICMP_STAT_BMCASTTSTAMP);
566 break;
567 }
568 icp->icmp_type = ICMP_TSTAMPREPLY;
569 icp->icmp_rtime = iptime();
570 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
571 goto reflect;
572
573 case ICMP_MASKREQ: {
574 struct ifnet *rcvif;
575 int s, ss;
576 struct ifaddr *ifa = NULL;
577
578 if (icmpmaskrepl == 0)
579 break;
580 /*
581 * We are not able to respond with all ones broadcast
582 * unless we receive it over a point-to-point interface.
583 */
584 if (icmplen < ICMP_MASKLEN) {
585 ICMP_STATINC(ICMP_STAT_BADLEN);
586 break;
587 }
588 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
589 in_nullhost(ip->ip_dst))
590 icmpdst.sin_addr = ip->ip_src;
591 else
592 icmpdst.sin_addr = ip->ip_dst;
593 ss = pserialize_read_enter();
594 rcvif = m_get_rcvif(m, &s);
595 if (__predict_true(rcvif != NULL))
596 ifa = ifaof_ifpforaddr(sintosa(&icmpdst), rcvif);
597 m_put_rcvif(rcvif, &s);
598 if (ifa == NULL) {
599 pserialize_read_exit(ss);
600 break;
601 }
602 ia = ifatoia(ifa);
603 icp->icmp_type = ICMP_MASKREPLY;
604 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
605 if (in_nullhost(ip->ip_src)) {
606 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
607 ip->ip_src = ia->ia_broadaddr.sin_addr;
608 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
609 ip->ip_src = ia->ia_dstaddr.sin_addr;
610 }
611 pserialize_read_exit(ss);
612 reflect:
613 {
614 uint64_t *icps = percpu_getref(icmpstat_percpu);
615 icps[ICMP_STAT_REFLECT]++;
616 icps[ICMP_STAT_OUTHIST + icp->icmp_type]++;
617 percpu_putref(icmpstat_percpu);
618 }
619 icmp_reflect(m);
620 return;
621 }
622
623 case ICMP_REDIRECT:
624 if (code > 3)
625 goto badcode;
626 if (icmp_rediraccept == 0)
627 goto freeit;
628 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
629 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
630 ICMP_STATINC(ICMP_STAT_BADLEN);
631 break;
632 }
633 /*
634 * Short circuit routing redirects to force
635 * immediate change in the kernel's routing
636 * tables. The message is also handed to anyone
637 * listening on a raw socket (e.g. the routing
638 * daemon for use in updating its tables).
639 */
640 icmpgw.sin_addr = ip->ip_src;
641 icmpdst.sin_addr = icp->icmp_gwaddr;
642 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
643 rt = NULL;
644 rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
645 NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt);
646 mutex_enter(&icmp_mtx);
647 if (rt != NULL && icmp_redirtimeout != 0) {
648 i = rt_timer_add(rt, icmp_redirect_timeout,
649 icmp_redirect_timeout_q);
650 if (i) {
651 char buf[INET_ADDRSTRLEN];
652 log(LOG_ERR, "ICMP: redirect failed to "
653 "register timeout for route to %s, "
654 "code %d\n",
655 IN_PRINT(buf, &icp->icmp_ip.ip_dst), i);
656 }
657 }
658 mutex_exit(&icmp_mtx);
659 if (rt != NULL)
660 rt_unref(rt);
661
662 pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
663 #if defined(IPSEC)
664 if (ipsec_used)
665 key_sa_routechange((struct sockaddr *)&icmpsrc);
666 #endif
667 break;
668
669 /*
670 * No kernel processing for the following;
671 * just fall through to send to raw listener.
672 */
673 case ICMP_ECHOREPLY:
674 case ICMP_ROUTERADVERT:
675 case ICMP_ROUTERSOLICIT:
676 case ICMP_TSTAMPREPLY:
677 case ICMP_IREQREPLY:
678 case ICMP_MASKREPLY:
679 default:
680 break;
681 }
682
683 raw:
684 /*
685 * Currently, pim_input() is always called holding softnet_lock
686 * by ipintr()(!NET_MPSAFE) or PR_INPUT_WRAP()(NET_MPSAFE).
687 */
688 KASSERT(mutex_owned(softnet_lock));
689 rip_input(m, hlen, proto);
690 return;
691
692 freeit:
693 m_freem(m);
694 return;
695 }
696
697 void
icmp_input(struct mbuf * m,int off,int proto)698 icmp_input(struct mbuf *m, int off, int proto)
699 {
700 wqinput_input(icmp_wqinput, m, off, proto);
701 }
702
703 /*
704 * Reflect the ip packet back to the source
705 */
706 void
icmp_reflect(struct mbuf * m)707 icmp_reflect(struct mbuf *m)
708 {
709 struct ip *ip = mtod(m, struct ip *);
710 struct in_ifaddr *ia;
711 struct ifaddr *ifa;
712 struct sockaddr_in *sin;
713 struct in_addr t;
714 struct mbuf *opts = NULL;
715 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
716 struct ifnet *rcvif;
717 struct psref psref, psref_ia;
718 int s;
719 int bound;
720
721 bound = curlwp_bind();
722
723 if (!in_canforward(ip->ip_src) &&
724 ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
725 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
726 m_freem(m); /* Bad return address */
727 goto done; /* ip_output() will check for broadcast */
728 }
729 t = ip->ip_dst;
730 ip->ip_dst = ip->ip_src;
731
732 /*
733 * If the incoming packet was addressed directly to us, use
734 * dst as the src for the reply. Otherwise (broadcast or
735 * anonymous), use an address which corresponds to the
736 * incoming interface, with a preference for the address which
737 * corresponds to the route to the destination of the ICMP.
738 */
739
740 /* Look for packet addressed to us */
741 ia = in_get_ia_psref(t, &psref_ia);
742 if (ia && (ia->ia4_flags & IN_IFF_NOTREADY)) {
743 ia4_release(ia, &psref_ia);
744 ia = NULL;
745 }
746
747 rcvif = m_get_rcvif_psref(m, &psref);
748
749 /* look for packet sent to broadcast address */
750 if (ia == NULL && rcvif &&
751 (rcvif->if_flags & IFF_BROADCAST)) {
752 s = pserialize_read_enter();
753 IFADDR_READER_FOREACH(ifa, rcvif) {
754 if (ifa->ifa_addr->sa_family != AF_INET)
755 continue;
756 if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) {
757 ia = ifatoia(ifa);
758 if ((ia->ia4_flags & IN_IFF_NOTREADY) == 0)
759 break;
760 ia = NULL;
761 }
762 }
763 if (ia != NULL)
764 ia4_acquire(ia, &psref_ia);
765 pserialize_read_exit(s);
766 }
767
768 sin = ia ? &ia->ia_addr : NULL;
769
770 /*
771 * if the packet is addressed somewhere else, compute the
772 * source address for packets routed back to the source, and
773 * use that, if it's an address on the interface which
774 * received the packet
775 */
776 if (sin == NULL && rcvif) {
777 struct sockaddr_in sin_dst;
778 struct route icmproute;
779 int errornum;
780
781 sockaddr_in_init(&sin_dst, &ip->ip_dst, 0);
782 memset(&icmproute, 0, sizeof(icmproute));
783 errornum = 0;
784 ia = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum,
785 &psref_ia);
786 /* errornum is never used */
787 rtcache_free(&icmproute);
788 /* check to make sure sin is a source address on rcvif */
789 if (ia != NULL) {
790 sin = &ia->ia_addr;
791 t = sin->sin_addr;
792 sin = NULL;
793 ia4_release(ia, &psref_ia);
794 ia = in_get_ia_on_iface_psref(t, rcvif, &psref_ia);
795 if (ia != NULL)
796 sin = &ia->ia_addr;
797 }
798 }
799
800 /*
801 * if it was not addressed to us, but the route doesn't go out
802 * the source interface, pick an address on the source
803 * interface. This can happen when routing is asymmetric, or
804 * when the incoming packet was encapsulated
805 */
806 if (sin == NULL && rcvif) {
807 KASSERT(ia == NULL);
808 s = pserialize_read_enter();
809 IFADDR_READER_FOREACH(ifa, rcvif) {
810 if (ifa->ifa_addr->sa_family != AF_INET)
811 continue;
812 sin = &(ifatoia(ifa)->ia_addr);
813 ia = ifatoia(ifa);
814 ia4_acquire(ia, &psref_ia);
815 break;
816 }
817 pserialize_read_exit(s);
818 }
819
820 m_put_rcvif_psref(rcvif, &psref);
821
822 /*
823 * The following happens if the packet was not addressed to us,
824 * and was received on an interface with no IP address:
825 * We find the first AF_INET address on the first non-loopback
826 * interface.
827 */
828 if (sin == NULL) {
829 KASSERT(ia == NULL);
830 s = pserialize_read_enter();
831 IN_ADDRLIST_READER_FOREACH(ia) {
832 if (ia->ia_ifp->if_flags & IFF_LOOPBACK)
833 continue;
834 sin = &ia->ia_addr;
835 ia4_acquire(ia, &psref_ia);
836 break;
837 }
838 pserialize_read_exit(s);
839 }
840
841 /*
842 * If we still didn't find an address, punt. We could have an
843 * interface up (and receiving packets) with no address.
844 */
845 if (sin == NULL) {
846 KASSERT(ia == NULL);
847 m_freem(m);
848 goto done;
849 }
850
851 ip->ip_src = sin->sin_addr;
852 ip->ip_ttl = MAXTTL;
853
854 if (ia != NULL)
855 ia4_release(ia, &psref_ia);
856
857 if (optlen > 0) {
858 u_char *cp;
859 int opt, cnt;
860 u_int len;
861
862 /*
863 * Retrieve any source routing from the incoming packet;
864 * add on any record-route or timestamp options.
865 */
866 cp = (u_char *)(ip + 1);
867 if ((opts = ip_srcroute(m)) == NULL &&
868 (MGETHDR(opts, M_DONTWAIT, MT_HEADER))) {
869 MCLAIM(opts, m->m_owner);
870 opts->m_len = sizeof(struct in_addr);
871 *mtod(opts, struct in_addr *) = zeroin_addr;
872 }
873
874 if (opts) {
875 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
876 opt = cp[IPOPT_OPTVAL];
877 if (opt == IPOPT_EOL)
878 break;
879 if (opt == IPOPT_NOP)
880 len = 1;
881 else {
882 if (cnt < IPOPT_OLEN + sizeof(*cp))
883 break;
884 len = cp[IPOPT_OLEN];
885 if (len < IPOPT_OLEN + sizeof(*cp) ||
886 len > cnt)
887 break;
888 }
889
890 /* Overflows can't happen */
891 KASSERT(opts->m_len + len <= MHLEN);
892
893 if (opt == IPOPT_RR || opt == IPOPT_TS ||
894 opt == IPOPT_SECURITY) {
895 memmove(mtod(opts, char *) +
896 opts->m_len, cp, len);
897 opts->m_len += len;
898 }
899 }
900
901 /* Terminate & pad, if necessary */
902 if ((cnt = opts->m_len % 4) != 0) {
903 for (; cnt < 4; cnt++) {
904 *(mtod(opts, char *) + opts->m_len) =
905 IPOPT_EOL;
906 opts->m_len++;
907 }
908 }
909 }
910
911 /*
912 * Now strip out original options by copying rest of first
913 * mbuf's data back, and adjust the IP length.
914 */
915 ip->ip_len = htons(ntohs(ip->ip_len) - optlen);
916 ip->ip_hl = sizeof(struct ip) >> 2;
917 m->m_len -= optlen;
918 if (m->m_flags & M_PKTHDR)
919 m->m_pkthdr.len -= optlen;
920 optlen += sizeof(struct ip);
921 memmove(ip + 1, (char *)ip + optlen,
922 (unsigned)(m->m_len - sizeof(struct ip)));
923 }
924 m_tag_delete_chain(m);
925 m->m_flags &= ~(M_BCAST|M_MCAST);
926
927 /*
928 * Clear any in-bound checksum flags for this packet.
929 */
930 if (m->m_flags & M_PKTHDR)
931 m->m_pkthdr.csum_flags = 0;
932
933 icmp_send(m, opts);
934 done:
935 curlwp_bindx(bound);
936 if (opts)
937 (void)m_free(opts);
938 }
939
940 /*
941 * Send an icmp packet back to the ip level,
942 * after supplying a checksum.
943 */
944 static void
icmp_send(struct mbuf * m,struct mbuf * opts)945 icmp_send(struct mbuf *m, struct mbuf *opts)
946 {
947 struct ip *ip = mtod(m, struct ip *);
948 int hlen;
949 struct icmp *icp;
950
951 hlen = ip->ip_hl << 2;
952 m->m_data += hlen;
953 m->m_len -= hlen;
954 icp = mtod(m, struct icmp *);
955 icp->icmp_cksum = 0;
956 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
957 m->m_data -= hlen;
958 m->m_len += hlen;
959
960 (void)ip_output(m, opts, NULL, 0, NULL, NULL);
961 }
962
963 n_time
iptime(void)964 iptime(void)
965 {
966 struct timeval atv;
967 u_long t;
968
969 microtime(&atv);
970 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
971 return (htonl(t));
972 }
973
974 /*
975 * sysctl helper routine for net.inet.icmp.returndatabytes. ensures
976 * that the new value is in the correct range.
977 */
978 static int
sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS)979 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS)
980 {
981 int error, t;
982 struct sysctlnode node;
983
984 node = *rnode;
985 node.sysctl_data = &t;
986 t = icmpreturndatabytes;
987 error = sysctl_lookup(SYSCTLFN_CALL(&node));
988 if (error || newp == NULL)
989 return error;
990
991 if (t < 8 || t > 512)
992 return EINVAL;
993 icmpreturndatabytes = t;
994
995 return 0;
996 }
997
998 /*
999 * sysctl helper routine for net.inet.icmp.redirtimeout. ensures that
1000 * the given value is not less than zero and then resets the timeout
1001 * queue.
1002 */
1003 static int
sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS)1004 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS)
1005 {
1006 int error, tmp;
1007 struct sysctlnode node;
1008
1009 mutex_enter(&icmp_mtx);
1010
1011 node = *rnode;
1012 node.sysctl_data = &tmp;
1013 tmp = icmp_redirtimeout;
1014 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1015 if (error || newp == NULL)
1016 goto out;
1017 if (tmp < 0) {
1018 error = EINVAL;
1019 goto out;
1020 }
1021 icmp_redirtimeout = tmp;
1022
1023 /*
1024 * was it a *defined* side-effect that anyone even *reading*
1025 * this value causes these things to happen?
1026 */
1027 if (icmp_redirect_timeout_q != NULL) {
1028 if (icmp_redirtimeout == 0) {
1029 rt_timer_queue_destroy(icmp_redirect_timeout_q);
1030 icmp_redirect_timeout_q = NULL;
1031 } else {
1032 rt_timer_queue_change(icmp_redirect_timeout_q,
1033 icmp_redirtimeout);
1034 }
1035 } else if (icmp_redirtimeout > 0) {
1036 icmp_redirect_timeout_q =
1037 rt_timer_queue_create(icmp_redirtimeout);
1038 }
1039 error = 0;
1040 out:
1041 mutex_exit(&icmp_mtx);
1042 return error;
1043 }
1044
1045 static int
sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS)1046 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS)
1047 {
1048
1049 return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS));
1050 }
1051
1052 static void
sysctl_netinet_icmp_setup(struct sysctllog ** clog)1053 sysctl_netinet_icmp_setup(struct sysctllog **clog)
1054 {
1055
1056 sysctl_createv(clog, 0, NULL, NULL,
1057 CTLFLAG_PERMANENT,
1058 CTLTYPE_NODE, "inet", NULL,
1059 NULL, 0, NULL, 0,
1060 CTL_NET, PF_INET, CTL_EOL);
1061 sysctl_createv(clog, 0, NULL, NULL,
1062 CTLFLAG_PERMANENT,
1063 CTLTYPE_NODE, "icmp",
1064 SYSCTL_DESCR("ICMPv4 related settings"),
1065 NULL, 0, NULL, 0,
1066 CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL);
1067
1068 sysctl_createv(clog, 0, NULL, NULL,
1069 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1070 CTLTYPE_INT, "maskrepl",
1071 SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"),
1072 NULL, 0, &icmpmaskrepl, 0,
1073 CTL_NET, PF_INET, IPPROTO_ICMP,
1074 ICMPCTL_MASKREPL, CTL_EOL);
1075 sysctl_createv(clog, 0, NULL, NULL,
1076 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1077 CTLTYPE_INT, "returndatabytes",
1078 SYSCTL_DESCR("Number of bytes to return in an ICMP "
1079 "error message"),
1080 sysctl_net_inet_icmp_returndatabytes, 0,
1081 &icmpreturndatabytes, 0,
1082 CTL_NET, PF_INET, IPPROTO_ICMP,
1083 ICMPCTL_RETURNDATABYTES, CTL_EOL);
1084 sysctl_createv(clog, 0, NULL, NULL,
1085 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1086 CTLTYPE_INT, "errppslimit",
1087 SYSCTL_DESCR("Maximum number of outgoing ICMP error "
1088 "messages per second"),
1089 NULL, 0, &icmperrppslim, 0,
1090 CTL_NET, PF_INET, IPPROTO_ICMP,
1091 ICMPCTL_ERRPPSLIMIT, CTL_EOL);
1092 sysctl_createv(clog, 0, NULL, NULL,
1093 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1094 CTLTYPE_INT, "rediraccept",
1095 SYSCTL_DESCR("Accept ICMP_REDIRECT messages"),
1096 NULL, 0, &icmp_rediraccept, 0,
1097 CTL_NET, PF_INET, IPPROTO_ICMP,
1098 ICMPCTL_REDIRACCEPT, CTL_EOL);
1099 sysctl_createv(clog, 0, NULL, NULL,
1100 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1101 CTLTYPE_INT, "redirtimeout",
1102 SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated "
1103 "routes"),
1104 sysctl_net_inet_icmp_redirtimeout, 0,
1105 &icmp_redirtimeout, 0,
1106 CTL_NET, PF_INET, IPPROTO_ICMP,
1107 ICMPCTL_REDIRTIMEOUT, CTL_EOL);
1108 sysctl_createv(clog, 0, NULL, NULL,
1109 CTLFLAG_PERMANENT,
1110 CTLTYPE_STRUCT, "stats",
1111 SYSCTL_DESCR("ICMP statistics"),
1112 sysctl_net_inet_icmp_stats, 0, NULL, 0,
1113 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS,
1114 CTL_EOL);
1115 sysctl_createv(clog, 0, NULL, NULL,
1116 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1117 CTLTYPE_INT, "bmcastecho",
1118 SYSCTL_DESCR("Respond to ICMP_ECHO or ICMP_TIMESTAMP "
1119 "message to the broadcast or multicast"),
1120 NULL, 0, &icmpbmcastecho, 0,
1121 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_BMCASTECHO,
1122 CTL_EOL);
1123 sysctl_createv(clog, 0, NULL, NULL,
1124 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1125 CTLTYPE_BOOL, "dynamic_rt_msg",
1126 SYSCTL_DESCR("Send routing message for RTF_DYNAMIC"),
1127 NULL, 0, &icmp_dynamic_rt_msg, 0,
1128 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_DYNAMIC_RT_MSG,
1129 CTL_EOL);
1130 }
1131
1132 void
icmp_statinc(u_int stat)1133 icmp_statinc(u_int stat)
1134 {
1135
1136 KASSERT(stat < ICMP_NSTATS);
1137 ICMP_STATINC(stat);
1138 }
1139
1140 /* Table of common MTUs */
1141 static const u_int mtu_table[] = {
1142 65535, 65280, 32000, 17914, 9180, 8166,
1143 4352, 2002, 1492, 1006, 508, 296, 68, 0
1144 };
1145
1146 void
icmp_mtudisc(struct icmp * icp,struct in_addr faddr)1147 icmp_mtudisc(struct icmp *icp, struct in_addr faddr)
1148 {
1149 struct icmp_mtudisc_callback *mc;
1150 struct sockaddr *dst = sintosa(&icmpsrc);
1151 struct rtentry *rt;
1152 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */
1153 int error;
1154
1155 rt = rtalloc1(dst, 1);
1156 if (rt == NULL)
1157 return;
1158
1159 /* If we didn't get a host route, allocate one */
1160 if ((rt->rt_flags & RTF_HOST) == 0) {
1161 struct rtentry *nrt;
1162
1163 error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
1164 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
1165 if (error) {
1166 rt_unref(rt);
1167 return;
1168 }
1169 nrt->rt_rmx = rt->rt_rmx;
1170 rt_newmsg_dynamic(RTM_ADD, nrt);
1171 rt_unref(rt);
1172 rt = nrt;
1173 }
1174
1175 mutex_enter(&icmp_mtx);
1176 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
1177 mutex_exit(&icmp_mtx);
1178 if (error) {
1179 rt_unref(rt);
1180 return;
1181 }
1182
1183 if (mtu == 0) {
1184 int i = 0;
1185
1186 mtu = ntohs(icp->icmp_ip.ip_len);
1187 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */
1188 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
1189 mtu -= (icp->icmp_ip.ip_hl << 2);
1190
1191 /* If we still can't guess a value, try the route */
1192 if (mtu == 0) {
1193 mtu = rt->rt_rmx.rmx_mtu;
1194
1195 /* If no route mtu, default to the interface mtu */
1196 if (mtu == 0)
1197 mtu = rt->rt_ifp->if_mtu;
1198 }
1199
1200 for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++) {
1201 if (mtu > mtu_table[i]) {
1202 mtu = mtu_table[i];
1203 break;
1204 }
1205 }
1206 }
1207
1208 /*
1209 * XXX: RTV_MTU is overloaded, since the admin can set it
1210 * to turn off PMTU for a route, and the kernel can
1211 * set it to indicate a serious problem with PMTU
1212 * on a route. We should be using a separate flag
1213 * for the kernel to indicate this.
1214 */
1215
1216 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1217 if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
1218 rt->rt_rmx.rmx_locks |= RTV_MTU;
1219 else if (rt->rt_rmx.rmx_mtu > mtu ||
1220 rt->rt_rmx.rmx_mtu == 0) {
1221 ICMP_STATINC(ICMP_STAT_PMTUCHG);
1222 rt->rt_rmx.rmx_mtu = mtu;
1223 }
1224 }
1225
1226 if (rt != NULL)
1227 rt_unref(rt);
1228
1229 /*
1230 * Notify protocols that the MTU for this destination
1231 * has changed.
1232 */
1233 mutex_enter(&icmp_mtx);
1234 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
1235 mc = LIST_NEXT(mc, mc_list))
1236 (*mc->mc_func)(faddr);
1237 mutex_exit(&icmp_mtx);
1238 }
1239
1240 /*
1241 * Return the next larger or smaller MTU plateau (table from RFC 1191)
1242 * given current value MTU. If DIR is less than zero, a larger plateau
1243 * is returned; otherwise, a smaller value is returned.
1244 */
1245 u_int
ip_next_mtu(u_int mtu,int dir)1246 ip_next_mtu(u_int mtu, int dir) /* XXX unused */
1247 {
1248 int i;
1249
1250 for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) {
1251 if (mtu >= mtu_table[i])
1252 break;
1253 }
1254
1255 if (dir < 0) {
1256 if (i == 0) {
1257 return 0;
1258 } else {
1259 return mtu_table[i - 1];
1260 }
1261 } else {
1262 if (mtu_table[i] == 0) {
1263 return 0;
1264 } else if (mtu > mtu_table[i]) {
1265 return mtu_table[i];
1266 } else {
1267 return mtu_table[i + 1];
1268 }
1269 }
1270 }
1271
1272 static void
icmp_mtudisc_timeout(struct rtentry * rt,struct rttimer * r)1273 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1274 {
1275 struct rtentry *retrt;
1276
1277 KASSERT(rt != NULL);
1278 rt_assert_referenced(rt);
1279
1280 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1281 (RTF_DYNAMIC | RTF_HOST)) {
1282 rtrequest(RTM_DELETE, rt_getkey(rt),
1283 rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
1284 rt_newmsg_dynamic(RTM_DELETE, retrt);
1285 rt_unref(rt);
1286 rt_free(retrt);
1287 } else {
1288 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1289 rt->rt_rmx.rmx_mtu = 0;
1290 }
1291 }
1292 }
1293
1294 static void
icmp_redirect_timeout(struct rtentry * rt,struct rttimer * r)1295 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1296 {
1297 struct rtentry *retrt;
1298
1299 KASSERT(rt != NULL);
1300 rt_assert_referenced(rt);
1301
1302 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1303 (RTF_DYNAMIC | RTF_HOST)) {
1304 rtrequest(RTM_DELETE, rt_getkey(rt),
1305 rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
1306 rt_newmsg_dynamic(RTM_DELETE, retrt);
1307 rt_unref(rt);
1308 rt_free(retrt);
1309 }
1310 }
1311
1312 /*
1313 * Perform rate limit check.
1314 * Returns 0 if it is okay to send the icmp packet.
1315 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate
1316 * limitation.
1317 *
1318 * XXX per-destination/type check necessary?
1319 */
1320 int
icmp_ratelimit(const struct in_addr * dst,const int type,const int code)1321 icmp_ratelimit(const struct in_addr *dst, const int type,
1322 const int code)
1323 {
1324
1325 /* PPS limit */
1326 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count,
1327 icmperrppslim)) {
1328 /* The packet is subject to rate limit */
1329 return 1;
1330 }
1331
1332 /* okay to send */
1333 return 0;
1334 }
1335