FreeBSD/Linux Kernel Cross Reference
sys/netinet/ip_icmp.c
1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. 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 University 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 REGENTS 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 REGENTS 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 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
32 */
33
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD: releng/12.0/sys/netinet/ip_icmp.c 340267 2018-11-08 21:58:23Z gjb $");
36
37 #include "opt_inet.h"
38
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/mbuf.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/rmlock.h>
48 #include <sys/sysctl.h>
49 #include <sys/syslog.h>
50
51 #include <net/if.h>
52 #include <net/if_var.h>
53 #include <net/if_types.h>
54 #include <net/route.h>
55 #include <net/vnet.h>
56
57 #include <netinet/in.h>
58 #include <netinet/in_fib.h>
59 #include <netinet/in_pcb.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip.h>
63 #include <netinet/ip_icmp.h>
64 #include <netinet/ip_var.h>
65 #include <netinet/ip_options.h>
66 #include <netinet/sctp.h>
67 #include <netinet/tcp.h>
68 #include <netinet/tcp_var.h>
69 #include <netinet/tcpip.h>
70 #include <netinet/icmp_var.h>
71
72
73 #ifdef INET
74
75 #include <machine/in_cksum.h>
76
77 #include <security/mac/mac_framework.h>
78 #endif /* INET */
79
80 /*
81 * ICMP routines: error generation, receive packet processing, and
82 * routines to turnaround packets back to the originator, and
83 * host table maintenance routines.
84 */
85 VNET_DEFINE_STATIC(int, icmplim) = 200;
86 #define V_icmplim VNET(icmplim)
87 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
88 &VNET_NAME(icmplim), 0,
89 "Maximum number of ICMP responses per second");
90
91 VNET_DEFINE_STATIC(int, icmplim_output) = 1;
92 #define V_icmplim_output VNET(icmplim_output)
93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
94 &VNET_NAME(icmplim_output), 0,
95 "Enable logging of ICMP response rate limiting");
96
97 #ifdef INET
98 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
99 VNET_PCPUSTAT_SYSINIT(icmpstat);
100 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
101 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
102
103 #ifdef VIMAGE
104 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
105 #endif /* VIMAGE */
106
107 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
108 #define V_icmpmaskrepl VNET(icmpmaskrepl)
109 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
110 &VNET_NAME(icmpmaskrepl), 0,
111 "Reply to ICMP Address Mask Request packets");
112
113 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
114 #define V_icmpmaskfake VNET(icmpmaskfake)
115 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
116 &VNET_NAME(icmpmaskfake), 0,
117 "Fake reply to ICMP Address Mask Request packets");
118
119 VNET_DEFINE(int, drop_redirect) = 0;
120 #define V_drop_redirect VNET(drop_redirect)
121 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
122 &VNET_NAME(drop_redirect), 0,
123 "Ignore ICMP redirects");
124
125 VNET_DEFINE_STATIC(int, log_redirect) = 0;
126 #define V_log_redirect VNET(log_redirect)
127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
128 &VNET_NAME(log_redirect), 0,
129 "Log ICMP redirects to the console");
130
131 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
132 #define V_reply_src VNET(reply_src)
133 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
134 &VNET_NAME(reply_src), IFNAMSIZ,
135 "ICMP reply source for non-local packets");
136
137 VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
138 #define V_icmp_rfi VNET(icmp_rfi)
139 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
140 &VNET_NAME(icmp_rfi), 0,
141 "ICMP reply from incoming interface for non-local packets");
142 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
143 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
144 #define V_icmp_quotelen VNET(icmp_quotelen)
145 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
146 &VNET_NAME(icmp_quotelen), 0,
147 "Number of bytes from original packet to quote in ICMP reply");
148
149 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
150 #define V_icmpbmcastecho VNET(icmpbmcastecho)
151 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
152 &VNET_NAME(icmpbmcastecho), 0,
153 "Reply to multicast ICMP Echo Request and Timestamp packets");
154
155 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
156 #define V_icmptstamprepl VNET(icmptstamprepl)
157 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW,
158 &VNET_NAME(icmptstamprepl), 0,
159 "Respond to ICMP Timestamp packets");
160
161 #ifdef ICMPPRINTFS
162 int icmpprintfs = 0;
163 #endif
164
165 static void icmp_reflect(struct mbuf *);
166 static void icmp_send(struct mbuf *, struct mbuf *);
167
168 extern struct protosw inetsw[];
169
170 /*
171 * Kernel module interface for updating icmpstat. The argument is an index
172 * into icmpstat treated as an array of u_long. While this encodes the
173 * general layout of icmpstat into the caller, it doesn't encode its
174 * location, so that future changes to add, for example, per-CPU stats
175 * support won't cause binary compatibility problems for kernel modules.
176 */
177 void
178 kmod_icmpstat_inc(int statnum)
179 {
180
181 counter_u64_add(VNET(icmpstat)[statnum], 1);
182 }
183
184 /*
185 * Generate an error packet of type error
186 * in response to bad packet ip.
187 */
188 void
189 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
190 {
191 struct ip *oip, *nip;
192 struct icmp *icp;
193 struct mbuf *m;
194 unsigned icmplen, icmpelen, nlen, oiphlen;
195
196 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
197 __func__));
198
199 if (type != ICMP_REDIRECT)
200 ICMPSTAT_INC(icps_error);
201 /*
202 * Don't send error:
203 * if the original packet was encrypted.
204 * if not the first fragment of message.
205 * in response to a multicast or broadcast packet.
206 * if the old packet protocol was an ICMP error message.
207 */
208 if (n->m_flags & M_DECRYPTED)
209 goto freeit;
210 if (n->m_flags & (M_BCAST|M_MCAST))
211 goto freeit;
212
213 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
214 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
215 goto freeit;
216 oip = mtod(n, struct ip *);
217 oiphlen = oip->ip_hl << 2;
218 if (n->m_len < oiphlen + ICMP_MINLEN)
219 goto freeit;
220 #ifdef ICMPPRINTFS
221 if (icmpprintfs)
222 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
223 #endif
224 if (oip->ip_off & htons(~(IP_MF|IP_DF)))
225 goto freeit;
226 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
227 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
228 oiphlen))->icmp_type)) {
229 ICMPSTAT_INC(icps_oldicmp);
230 goto freeit;
231 }
232 /*
233 * Calculate length to quote from original packet and
234 * prevent the ICMP mbuf from overflowing.
235 * Unfortunately this is non-trivial since ip_forward()
236 * sends us truncated packets.
237 */
238 nlen = m_length(n, NULL);
239 if (oip->ip_p == IPPROTO_TCP) {
240 struct tcphdr *th;
241 int tcphlen;
242
243 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
244 n->m_next == NULL)
245 goto stdreply;
246 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
247 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
248 goto freeit;
249 oip = mtod(n, struct ip *);
250 th = mtodo(n, oiphlen);
251 tcphlen = th->th_off << 2;
252 if (tcphlen < sizeof(struct tcphdr))
253 goto freeit;
254 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
255 goto freeit;
256 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
257 goto stdreply;
258 if (n->m_len < oiphlen + tcphlen &&
259 (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
260 goto freeit;
261 icmpelen = max(tcphlen, min(V_icmp_quotelen,
262 ntohs(oip->ip_len) - oiphlen));
263 } else if (oip->ip_p == IPPROTO_SCTP) {
264 struct sctphdr *sh;
265 struct sctp_chunkhdr *ch;
266
267 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
268 goto stdreply;
269 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
270 n->m_next == NULL)
271 goto stdreply;
272 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
273 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
274 goto freeit;
275 oip = mtod(n, struct ip *);
276 icmpelen = max(sizeof(struct sctphdr),
277 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
278 sh = mtodo(n, oiphlen);
279 if (ntohl(sh->v_tag) == 0 &&
280 ntohs(oip->ip_len) >= oiphlen +
281 sizeof(struct sctphdr) + 8 &&
282 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
283 n->m_next != NULL)) {
284 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
285 (n = m_pullup(n, oiphlen +
286 sizeof(struct sctphdr) + 8)) == NULL)
287 goto freeit;
288 oip = mtod(n, struct ip *);
289 sh = mtodo(n, oiphlen);
290 ch = (struct sctp_chunkhdr *)(sh + 1);
291 if (ch->chunk_type == SCTP_INITIATION) {
292 icmpelen = max(sizeof(struct sctphdr) + 8,
293 min(V_icmp_quotelen, ntohs(oip->ip_len) -
294 oiphlen));
295 }
296 }
297 } else
298 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
299 oiphlen));
300
301 icmplen = min(oiphlen + icmpelen, nlen);
302 if (icmplen < sizeof(struct ip))
303 goto freeit;
304
305 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
306 m = m_gethdr(M_NOWAIT, MT_DATA);
307 else
308 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
309 if (m == NULL)
310 goto freeit;
311 #ifdef MAC
312 mac_netinet_icmp_reply(n, m);
313 #endif
314 icmplen = min(icmplen, M_TRAILINGSPACE(m) -
315 sizeof(struct ip) - ICMP_MINLEN);
316 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
317 m->m_data += sizeof(struct ip);
318 m->m_len = ICMP_MINLEN + icmplen;
319
320 /* XXX MRT make the outgoing packet use the same FIB
321 * that was associated with the incoming packet
322 */
323 M_SETFIB(m, M_GETFIB(n));
324 icp = mtod(m, struct icmp *);
325 ICMPSTAT_INC(icps_outhist[type]);
326 icp->icmp_type = type;
327 if (type == ICMP_REDIRECT)
328 icp->icmp_gwaddr.s_addr = dest;
329 else {
330 icp->icmp_void = 0;
331 /*
332 * The following assignments assume an overlay with the
333 * just zeroed icmp_void field.
334 */
335 if (type == ICMP_PARAMPROB) {
336 icp->icmp_pptr = code;
337 code = 0;
338 } else if (type == ICMP_UNREACH &&
339 code == ICMP_UNREACH_NEEDFRAG && mtu) {
340 icp->icmp_nextmtu = htons(mtu);
341 }
342 }
343 icp->icmp_code = code;
344
345 /*
346 * Copy the quotation into ICMP message and
347 * convert quoted IP header back to network representation.
348 */
349 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
350 nip = &icp->icmp_ip;
351
352 /*
353 * Set up ICMP message mbuf and copy old IP header (without options
354 * in front of ICMP message.
355 * If the original mbuf was meant to bypass the firewall, the error
356 * reply should bypass as well.
357 */
358 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
359 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
360 ("insufficient space for ip header"));
361 m->m_data -= sizeof(struct ip);
362 m->m_len += sizeof(struct ip);
363 m->m_pkthdr.len = m->m_len;
364 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
365 nip = mtod(m, struct ip *);
366 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
367 nip->ip_len = htons(m->m_len);
368 nip->ip_v = IPVERSION;
369 nip->ip_hl = 5;
370 nip->ip_p = IPPROTO_ICMP;
371 nip->ip_tos = 0;
372 nip->ip_off = 0;
373 icmp_reflect(m);
374
375 freeit:
376 m_freem(n);
377 }
378
379 /*
380 * Process a received ICMP message.
381 */
382 int
383 icmp_input(struct mbuf **mp, int *offp, int proto)
384 {
385 struct icmp *icp;
386 struct in_ifaddr *ia;
387 struct mbuf *m = *mp;
388 struct ip *ip = mtod(m, struct ip *);
389 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
390 int hlen = *offp;
391 int icmplen = ntohs(ip->ip_len) - *offp;
392 int i, code;
393 void (*ctlfunc)(int, struct sockaddr *, void *);
394 int fibnum;
395
396 *mp = NULL;
397
398 /*
399 * Locate icmp structure in mbuf, and check
400 * that not corrupted and of at least minimum length.
401 */
402 #ifdef ICMPPRINTFS
403 if (icmpprintfs) {
404 char srcbuf[INET_ADDRSTRLEN];
405 char dstbuf[INET_ADDRSTRLEN];
406
407 printf("icmp_input from %s to %s, len %d\n",
408 inet_ntoa_r(ip->ip_src, srcbuf),
409 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
410 }
411 #endif
412 NET_EPOCH_ENTER();
413 if (icmplen < ICMP_MINLEN) {
414 ICMPSTAT_INC(icps_tooshort);
415 goto freeit;
416 }
417 i = hlen + min(icmplen, ICMP_ADVLENMIN);
418 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
419 ICMPSTAT_INC(icps_tooshort);
420 NET_EPOCH_EXIT();
421 return (IPPROTO_DONE);
422 }
423 ip = mtod(m, struct ip *);
424 m->m_len -= hlen;
425 m->m_data += hlen;
426 icp = mtod(m, struct icmp *);
427 if (in_cksum(m, icmplen)) {
428 ICMPSTAT_INC(icps_checksum);
429 goto freeit;
430 }
431 m->m_len += hlen;
432 m->m_data -= hlen;
433
434 #ifdef ICMPPRINTFS
435 if (icmpprintfs)
436 printf("icmp_input, type %d code %d\n", icp->icmp_type,
437 icp->icmp_code);
438 #endif
439
440 /*
441 * Message type specific processing.
442 */
443 if (icp->icmp_type > ICMP_MAXTYPE)
444 goto raw;
445
446 /* Initialize */
447 bzero(&icmpsrc, sizeof(icmpsrc));
448 icmpsrc.sin_len = sizeof(struct sockaddr_in);
449 icmpsrc.sin_family = AF_INET;
450 bzero(&icmpdst, sizeof(icmpdst));
451 icmpdst.sin_len = sizeof(struct sockaddr_in);
452 icmpdst.sin_family = AF_INET;
453 bzero(&icmpgw, sizeof(icmpgw));
454 icmpgw.sin_len = sizeof(struct sockaddr_in);
455 icmpgw.sin_family = AF_INET;
456
457 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
458 code = icp->icmp_code;
459 switch (icp->icmp_type) {
460
461 case ICMP_UNREACH:
462 switch (code) {
463 case ICMP_UNREACH_NET:
464 case ICMP_UNREACH_HOST:
465 case ICMP_UNREACH_SRCFAIL:
466 case ICMP_UNREACH_NET_UNKNOWN:
467 case ICMP_UNREACH_HOST_UNKNOWN:
468 case ICMP_UNREACH_ISOLATED:
469 case ICMP_UNREACH_TOSNET:
470 case ICMP_UNREACH_TOSHOST:
471 case ICMP_UNREACH_HOST_PRECEDENCE:
472 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
473 code = PRC_UNREACH_NET;
474 break;
475
476 case ICMP_UNREACH_NEEDFRAG:
477 code = PRC_MSGSIZE;
478 break;
479
480 /*
481 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
482 * Treat subcodes 2,3 as immediate RST
483 */
484 case ICMP_UNREACH_PROTOCOL:
485 code = PRC_UNREACH_PROTOCOL;
486 break;
487 case ICMP_UNREACH_PORT:
488 code = PRC_UNREACH_PORT;
489 break;
490
491 case ICMP_UNREACH_NET_PROHIB:
492 case ICMP_UNREACH_HOST_PROHIB:
493 case ICMP_UNREACH_FILTER_PROHIB:
494 code = PRC_UNREACH_ADMIN_PROHIB;
495 break;
496
497 default:
498 goto badcode;
499 }
500 goto deliver;
501
502 case ICMP_TIMXCEED:
503 if (code > 1)
504 goto badcode;
505 code += PRC_TIMXCEED_INTRANS;
506 goto deliver;
507
508 case ICMP_PARAMPROB:
509 if (code > 1)
510 goto badcode;
511 code = PRC_PARAMPROB;
512 deliver:
513 /*
514 * Problem with datagram; advise higher level routines.
515 */
516 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
517 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
518 ICMPSTAT_INC(icps_badlen);
519 goto freeit;
520 }
521 /* Discard ICMP's in response to multicast packets */
522 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
523 goto badcode;
524 #ifdef ICMPPRINTFS
525 if (icmpprintfs)
526 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
527 #endif
528 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
529 /*
530 * XXX if the packet contains [IPv4 AH TCP], we can't make a
531 * notification to TCP layer.
532 */
533 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
534 ip_stripoptions(m);
535 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
536 /* This should actually not happen */
537 ICMPSTAT_INC(icps_tooshort);
538 NET_EPOCH_EXIT();
539 return (IPPROTO_DONE);
540 }
541 ip = mtod(m, struct ip *);
542 icp = (struct icmp *)(ip + 1);
543 /*
544 * The upper layer handler can rely on:
545 * - The outer IP header has no options.
546 * - The outer IP header, the ICMP header, the inner IP header,
547 * and the first n bytes of the inner payload are contiguous.
548 * n is at least 8, but might be larger based on
549 * ICMP_ADVLENPREF. See its definition in ip_icmp.h.
550 */
551 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
552 if (ctlfunc)
553 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
554 (void *)&icp->icmp_ip);
555 break;
556
557 badcode:
558 ICMPSTAT_INC(icps_badcode);
559 break;
560
561 case ICMP_ECHO:
562 if (!V_icmpbmcastecho
563 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
564 ICMPSTAT_INC(icps_bmcastecho);
565 break;
566 }
567 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
568 goto freeit;
569 icp->icmp_type = ICMP_ECHOREPLY;
570 goto reflect;
571
572 case ICMP_TSTAMP:
573 if (V_icmptstamprepl == 0)
574 break;
575 if (!V_icmpbmcastecho
576 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
577 ICMPSTAT_INC(icps_bmcasttstamp);
578 break;
579 }
580 if (icmplen < ICMP_TSLEN) {
581 ICMPSTAT_INC(icps_badlen);
582 break;
583 }
584 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
585 goto freeit;
586 icp->icmp_type = ICMP_TSTAMPREPLY;
587 icp->icmp_rtime = iptime();
588 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
589 goto reflect;
590
591 case ICMP_MASKREQ:
592 if (V_icmpmaskrepl == 0)
593 break;
594 /*
595 * We are not able to respond with all ones broadcast
596 * unless we receive it over a point-to-point interface.
597 */
598 if (icmplen < ICMP_MASKLEN)
599 break;
600 switch (ip->ip_dst.s_addr) {
601
602 case INADDR_BROADCAST:
603 case INADDR_ANY:
604 icmpdst.sin_addr = ip->ip_src;
605 break;
606
607 default:
608 icmpdst.sin_addr = ip->ip_dst;
609 }
610 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
611 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
612 if (ia == NULL)
613 break;
614 if (ia->ia_ifp == NULL)
615 break;
616 icp->icmp_type = ICMP_MASKREPLY;
617 if (V_icmpmaskfake == 0)
618 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
619 else
620 icp->icmp_mask = V_icmpmaskfake;
621 if (ip->ip_src.s_addr == 0) {
622 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
623 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
624 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
625 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
626 }
627 reflect:
628 ICMPSTAT_INC(icps_reflect);
629 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
630 icmp_reflect(m);
631 NET_EPOCH_EXIT();
632 return (IPPROTO_DONE);
633
634 case ICMP_REDIRECT:
635 if (V_log_redirect) {
636 u_long src, dst, gw;
637
638 src = ntohl(ip->ip_src.s_addr);
639 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
640 gw = ntohl(icp->icmp_gwaddr.s_addr);
641 printf("icmp redirect from %d.%d.%d.%d: "
642 "%d.%d.%d.%d => %d.%d.%d.%d\n",
643 (int)(src >> 24), (int)((src >> 16) & 0xff),
644 (int)((src >> 8) & 0xff), (int)(src & 0xff),
645 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
646 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
647 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
648 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
649 }
650 /*
651 * RFC1812 says we must ignore ICMP redirects if we
652 * are acting as router.
653 */
654 if (V_drop_redirect || V_ipforwarding)
655 break;
656 if (code > 3)
657 goto badcode;
658 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
659 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
660 ICMPSTAT_INC(icps_badlen);
661 break;
662 }
663 /*
664 * Short circuit routing redirects to force
665 * immediate change in the kernel's routing
666 * tables. The message is also handed to anyone
667 * listening on a raw socket (e.g. the routing
668 * daemon for use in updating its tables).
669 */
670 icmpgw.sin_addr = ip->ip_src;
671 icmpdst.sin_addr = icp->icmp_gwaddr;
672 #ifdef ICMPPRINTFS
673 if (icmpprintfs) {
674 char dstbuf[INET_ADDRSTRLEN];
675 char gwbuf[INET_ADDRSTRLEN];
676
677 printf("redirect dst %s to %s\n",
678 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
679 inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
680 }
681 #endif
682 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
683 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
684 in_rtredirect((struct sockaddr *)&icmpsrc,
685 (struct sockaddr *)&icmpdst,
686 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
687 (struct sockaddr *)&icmpgw, fibnum);
688 }
689 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
690 break;
691
692 /*
693 * No kernel processing for the following;
694 * just fall through to send to raw listener.
695 */
696 case ICMP_ECHOREPLY:
697 case ICMP_ROUTERADVERT:
698 case ICMP_ROUTERSOLICIT:
699 case ICMP_TSTAMPREPLY:
700 case ICMP_IREQREPLY:
701 case ICMP_MASKREPLY:
702 case ICMP_SOURCEQUENCH:
703 default:
704 break;
705 }
706
707 raw:
708 NET_EPOCH_EXIT();
709 *mp = m;
710 rip_input(mp, offp, proto);
711 return (IPPROTO_DONE);
712
713 freeit:
714 NET_EPOCH_EXIT();
715 m_freem(m);
716 return (IPPROTO_DONE);
717 }
718
719 /*
720 * Reflect the ip packet back to the source
721 */
722 static void
723 icmp_reflect(struct mbuf *m)
724 {
725 struct rm_priotracker in_ifa_tracker;
726 struct ip *ip = mtod(m, struct ip *);
727 struct ifaddr *ifa;
728 struct ifnet *ifp;
729 struct in_ifaddr *ia;
730 struct in_addr t;
731 struct nhop4_extended nh_ext;
732 struct mbuf *opts = NULL;
733 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
734
735 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
736 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
737 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
738 m_freem(m); /* Bad return address */
739 ICMPSTAT_INC(icps_badaddr);
740 goto done; /* Ip_output() will check for broadcast */
741 }
742
743 t = ip->ip_dst;
744 ip->ip_dst = ip->ip_src;
745
746 /*
747 * Source selection for ICMP replies:
748 *
749 * If the incoming packet was addressed directly to one of our
750 * own addresses, use dst as the src for the reply.
751 */
752 IN_IFADDR_RLOCK(&in_ifa_tracker);
753 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
754 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
755 t = IA_SIN(ia)->sin_addr;
756 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
757 goto match;
758 }
759 }
760 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
761
762 /*
763 * If the incoming packet was addressed to one of our broadcast
764 * addresses, use the first non-broadcast address which corresponds
765 * to the incoming interface.
766 */
767 ifp = m->m_pkthdr.rcvif;
768 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
769 IF_ADDR_RLOCK(ifp);
770 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
771 if (ifa->ifa_addr->sa_family != AF_INET)
772 continue;
773 ia = ifatoia(ifa);
774 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
775 t.s_addr) {
776 t = IA_SIN(ia)->sin_addr;
777 IF_ADDR_RUNLOCK(ifp);
778 goto match;
779 }
780 }
781 IF_ADDR_RUNLOCK(ifp);
782 }
783 /*
784 * If the packet was transiting through us, use the address of
785 * the interface the packet came through in. If that interface
786 * doesn't have a suitable IP address, the normal selection
787 * criteria apply.
788 */
789 if (V_icmp_rfi && ifp != NULL) {
790 IF_ADDR_RLOCK(ifp);
791 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
792 if (ifa->ifa_addr->sa_family != AF_INET)
793 continue;
794 ia = ifatoia(ifa);
795 t = IA_SIN(ia)->sin_addr;
796 IF_ADDR_RUNLOCK(ifp);
797 goto match;
798 }
799 IF_ADDR_RUNLOCK(ifp);
800 }
801 /*
802 * If the incoming packet was not addressed directly to us, use
803 * designated interface for icmp replies specified by sysctl
804 * net.inet.icmp.reply_src (default not set). Otherwise continue
805 * with normal source selection.
806 */
807 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
808 IF_ADDR_RLOCK(ifp);
809 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
810 if (ifa->ifa_addr->sa_family != AF_INET)
811 continue;
812 ia = ifatoia(ifa);
813 t = IA_SIN(ia)->sin_addr;
814 IF_ADDR_RUNLOCK(ifp);
815 goto match;
816 }
817 IF_ADDR_RUNLOCK(ifp);
818 }
819 /*
820 * If the packet was transiting through us, use the address of
821 * the interface that is the closest to the packet source.
822 * When we don't have a route back to the packet source, stop here
823 * and drop the packet.
824 */
825 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
826 m_freem(m);
827 ICMPSTAT_INC(icps_noroute);
828 goto done;
829 }
830 t = nh_ext.nh_src;
831 match:
832 #ifdef MAC
833 mac_netinet_icmp_replyinplace(m);
834 #endif
835 ip->ip_src = t;
836 ip->ip_ttl = V_ip_defttl;
837
838 if (optlen > 0) {
839 u_char *cp;
840 int opt, cnt;
841 u_int len;
842
843 /*
844 * Retrieve any source routing from the incoming packet;
845 * add on any record-route or timestamp options.
846 */
847 cp = (u_char *) (ip + 1);
848 if ((opts = ip_srcroute(m)) == NULL &&
849 (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
850 opts->m_len = sizeof(struct in_addr);
851 mtod(opts, struct in_addr *)->s_addr = 0;
852 }
853 if (opts) {
854 #ifdef ICMPPRINTFS
855 if (icmpprintfs)
856 printf("icmp_reflect optlen %d rt %d => ",
857 optlen, opts->m_len);
858 #endif
859 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
860 opt = cp[IPOPT_OPTVAL];
861 if (opt == IPOPT_EOL)
862 break;
863 if (opt == IPOPT_NOP)
864 len = 1;
865 else {
866 if (cnt < IPOPT_OLEN + sizeof(*cp))
867 break;
868 len = cp[IPOPT_OLEN];
869 if (len < IPOPT_OLEN + sizeof(*cp) ||
870 len > cnt)
871 break;
872 }
873 /*
874 * Should check for overflow, but it "can't happen"
875 */
876 if (opt == IPOPT_RR || opt == IPOPT_TS ||
877 opt == IPOPT_SECURITY) {
878 bcopy((caddr_t)cp,
879 mtod(opts, caddr_t) + opts->m_len, len);
880 opts->m_len += len;
881 }
882 }
883 /* Terminate & pad, if necessary */
884 cnt = opts->m_len % 4;
885 if (cnt) {
886 for (; cnt < 4; cnt++) {
887 *(mtod(opts, caddr_t) + opts->m_len) =
888 IPOPT_EOL;
889 opts->m_len++;
890 }
891 }
892 #ifdef ICMPPRINTFS
893 if (icmpprintfs)
894 printf("%d\n", opts->m_len);
895 #endif
896 }
897 ip_stripoptions(m);
898 }
899 m_tag_delete_nonpersistent(m);
900 m->m_flags &= ~(M_BCAST|M_MCAST);
901 icmp_send(m, opts);
902 done:
903 if (opts)
904 (void)m_free(opts);
905 }
906
907 /*
908 * Send an icmp packet back to the ip level,
909 * after supplying a checksum.
910 */
911 static void
912 icmp_send(struct mbuf *m, struct mbuf *opts)
913 {
914 struct ip *ip = mtod(m, struct ip *);
915 int hlen;
916 struct icmp *icp;
917
918 hlen = ip->ip_hl << 2;
919 m->m_data += hlen;
920 m->m_len -= hlen;
921 icp = mtod(m, struct icmp *);
922 icp->icmp_cksum = 0;
923 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
924 m->m_data -= hlen;
925 m->m_len += hlen;
926 m->m_pkthdr.rcvif = (struct ifnet *)0;
927 #ifdef ICMPPRINTFS
928 if (icmpprintfs) {
929 char dstbuf[INET_ADDRSTRLEN];
930 char srcbuf[INET_ADDRSTRLEN];
931
932 printf("icmp_send dst %s src %s\n",
933 inet_ntoa_r(ip->ip_dst, dstbuf),
934 inet_ntoa_r(ip->ip_src, srcbuf));
935 }
936 #endif
937 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
938 }
939
940 /*
941 * Return milliseconds since 00:00 UTC in network format.
942 */
943 uint32_t
944 iptime(void)
945 {
946 struct timeval atv;
947 u_long t;
948
949 getmicrotime(&atv);
950 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
951 return (htonl(t));
952 }
953
954 /*
955 * Return the next larger or smaller MTU plateau (table from RFC 1191)
956 * given current value MTU. If DIR is less than zero, a larger plateau
957 * is returned; otherwise, a smaller value is returned.
958 */
959 int
960 ip_next_mtu(int mtu, int dir)
961 {
962 static int mtutab[] = {
963 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
964 296, 68, 0
965 };
966 int i, size;
967
968 size = (sizeof mtutab) / (sizeof mtutab[0]);
969 if (dir >= 0) {
970 for (i = 0; i < size; i++)
971 if (mtu > mtutab[i])
972 return mtutab[i];
973 } else {
974 for (i = size - 1; i >= 0; i--)
975 if (mtu < mtutab[i])
976 return mtutab[i];
977 if (mtu == mtutab[0])
978 return mtutab[0];
979 }
980 return 0;
981 }
982 #endif /* INET */
983
984
985 /*
986 * badport_bandlim() - check for ICMP bandwidth limit
987 *
988 * Return 0 if it is ok to send an ICMP error response, -1 if we have
989 * hit our bandwidth limit and it is not ok.
990 *
991 * If icmplim is <= 0, the feature is disabled and 0 is returned.
992 *
993 * For now we separate the TCP and UDP subsystems w/ different 'which'
994 * values. We may eventually remove this separation (and simplify the
995 * code further).
996 *
997 * Note that the printing of the error message is delayed so we can
998 * properly print the icmp error rate that the system was trying to do
999 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
1000 * the 'final' error, but it doesn't make sense to solve the printing
1001 * delay with more complex code.
1002 */
1003 struct icmp_rate {
1004 const char *descr;
1005 struct counter_rate cr;
1006 };
1007 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
1008 { "icmp unreach response" },
1009 { "icmp ping response" },
1010 { "icmp tstamp response" },
1011 { "closed port RST response" },
1012 { "open port RST response" },
1013 { "icmp6 unreach response" },
1014 { "sctp ootb response" }
1015 };
1016 #define V_icmp_rates VNET(icmp_rates)
1017
1018 static void
1019 icmp_bandlimit_init(void)
1020 {
1021
1022 for (int i = 0; i < BANDLIM_MAX; i++) {
1023 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1024 V_icmp_rates[i].cr.cr_ticks = ticks;
1025 }
1026 }
1027 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1028 icmp_bandlimit_init, NULL);
1029
1030 static void
1031 icmp_bandlimit_uninit(void)
1032 {
1033
1034 for (int i = 0; i < BANDLIM_MAX; i++)
1035 counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1036 }
1037 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1038 icmp_bandlimit_uninit, NULL);
1039
1040 int
1041 badport_bandlim(int which)
1042 {
1043 int64_t pps;
1044
1045 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1046 return (0);
1047
1048 KASSERT(which >= 0 && which < BANDLIM_MAX,
1049 ("%s: which %d", __func__, which));
1050
1051 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim);
1052 if (pps == -1)
1053 return (-1);
1054 if (pps > 0 && V_icmplim_output)
1055 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1056 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim);
1057 return (0);
1058 }
Cache object: 89cd6453fb945b74f9f85f54435dfc15
|