xref: /freebsd/sys/netinet/ip_icmp.c (revision ca987d4641cdcd7f27e153db17c5bf064934faf5)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/rmlock.h>
46 #include <sys/sysctl.h>
47 #include <sys/syslog.h>
48 
49 #include <net/if.h>
50 #include <net/if_var.h>
51 #include <net/if_types.h>
52 #include <net/route.h>
53 #include <net/vnet.h>
54 
55 #include <netinet/in.h>
56 #include <netinet/in_fib.h>
57 #include <netinet/in_pcb.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_icmp.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_options.h>
64 #include <netinet/sctp.h>
65 #include <netinet/tcp.h>
66 #include <netinet/tcp_var.h>
67 #include <netinet/tcpip.h>
68 #include <netinet/icmp_var.h>
69 
70 #ifdef INET
71 
72 #include <machine/in_cksum.h>
73 
74 #include <security/mac/mac_framework.h>
75 #endif /* INET */
76 
77 /*
78  * ICMP routines: error generation, receive packet processing, and
79  * routines to turnaround packets back to the originator, and
80  * host table maintenance routines.
81  */
82 static VNET_DEFINE(int, icmplim) = 200;
83 #define	V_icmplim			VNET(icmplim)
84 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
85 	&VNET_NAME(icmplim), 0,
86 	"Maximum number of ICMP responses per second");
87 
88 static VNET_DEFINE(int, icmplim_output) = 1;
89 #define	V_icmplim_output		VNET(icmplim_output)
90 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
91 	&VNET_NAME(icmplim_output), 0,
92 	"Enable logging of ICMP response rate limiting");
93 
94 #ifdef INET
95 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
96 VNET_PCPUSTAT_SYSINIT(icmpstat);
97 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
98     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
99 
100 #ifdef VIMAGE
101 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
102 #endif /* VIMAGE */
103 
104 static VNET_DEFINE(int, icmpmaskrepl) = 0;
105 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
106 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
107 	&VNET_NAME(icmpmaskrepl), 0,
108 	"Reply to ICMP Address Mask Request packets");
109 
110 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
111 #define	V_icmpmaskfake			VNET(icmpmaskfake)
112 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
113 	&VNET_NAME(icmpmaskfake), 0,
114 	"Fake reply to ICMP Address Mask Request packets");
115 
116 VNET_DEFINE(int, drop_redirect) = 0;
117 #define	V_drop_redirect			VNET(drop_redirect)
118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
119 	&VNET_NAME(drop_redirect), 0,
120 	"Ignore ICMP redirects");
121 
122 static VNET_DEFINE(int, log_redirect) = 0;
123 #define	V_log_redirect			VNET(log_redirect)
124 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
125 	&VNET_NAME(log_redirect), 0,
126 	"Log ICMP redirects to the console");
127 
128 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
129 #define	V_reply_src			VNET(reply_src)
130 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
131 	&VNET_NAME(reply_src), IFNAMSIZ,
132 	"ICMP reply source for non-local packets");
133 
134 static VNET_DEFINE(int, icmp_rfi) = 0;
135 #define	V_icmp_rfi			VNET(icmp_rfi)
136 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
137 	&VNET_NAME(icmp_rfi), 0,
138 	"ICMP reply from incoming interface for non-local packets");
139 
140 static VNET_DEFINE(int, icmp_quotelen) = 8;
141 #define	V_icmp_quotelen			VNET(icmp_quotelen)
142 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
143 	&VNET_NAME(icmp_quotelen), 0,
144 	"Number of bytes from original packet to quote in ICMP reply");
145 
146 static VNET_DEFINE(int, icmpbmcastecho) = 0;
147 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
148 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
149 	&VNET_NAME(icmpbmcastecho), 0,
150 	"Reply to multicast ICMP Echo Request and Timestamp packets");
151 
152 static VNET_DEFINE(int, icmptstamprepl) = 1;
153 #define	V_icmptstamprepl		VNET(icmptstamprepl)
154 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW,
155 	&VNET_NAME(icmptstamprepl), 0,
156 	"Respond to ICMP Timestamp packets");
157 
158 #ifdef ICMPPRINTFS
159 int	icmpprintfs = 0;
160 #endif
161 
162 static void	icmp_reflect(struct mbuf *);
163 static void	icmp_send(struct mbuf *, struct mbuf *);
164 
165 extern	struct protosw inetsw[];
166 
167 /*
168  * Kernel module interface for updating icmpstat.  The argument is an index
169  * into icmpstat treated as an array of u_long.  While this encodes the
170  * general layout of icmpstat into the caller, it doesn't encode its
171  * location, so that future changes to add, for example, per-CPU stats
172  * support won't cause binary compatibility problems for kernel modules.
173  */
174 void
175 kmod_icmpstat_inc(int statnum)
176 {
177 
178 	counter_u64_add(VNET(icmpstat)[statnum], 1);
179 }
180 
181 /*
182  * Generate an error packet of type error
183  * in response to bad packet ip.
184  */
185 void
186 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
187 {
188 	struct ip *oip, *nip;
189 	struct icmp *icp;
190 	struct mbuf *m;
191 	unsigned icmplen, icmpelen, nlen, oiphlen;
192 
193 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
194 	    __func__));
195 
196 	if (type != ICMP_REDIRECT)
197 		ICMPSTAT_INC(icps_error);
198 	/*
199 	 * Don't send error:
200 	 *  if the original packet was encrypted.
201 	 *  if not the first fragment of message.
202 	 *  in response to a multicast or broadcast packet.
203 	 *  if the old packet protocol was an ICMP error message.
204 	 */
205 	if (n->m_flags & M_DECRYPTED)
206 		goto freeit;
207 	if (n->m_flags & (M_BCAST|M_MCAST))
208 		goto freeit;
209 
210 	/* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
211 	if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
212 		goto freeit;
213 	oip = mtod(n, struct ip *);
214 	oiphlen = oip->ip_hl << 2;
215 	if (n->m_len < oiphlen + ICMP_MINLEN)
216 		goto freeit;
217 #ifdef ICMPPRINTFS
218 	if (icmpprintfs)
219 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
220 #endif
221 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
222 		goto freeit;
223 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
224 	    !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
225 		oiphlen))->icmp_type)) {
226 		ICMPSTAT_INC(icps_oldicmp);
227 		goto freeit;
228 	}
229 	/*
230 	 * Calculate length to quote from original packet and
231 	 * prevent the ICMP mbuf from overflowing.
232 	 * Unfortunately this is non-trivial since ip_forward()
233 	 * sends us truncated packets.
234 	 */
235 	nlen = m_length(n, NULL);
236 	if (oip->ip_p == IPPROTO_TCP) {
237 		struct tcphdr *th;
238 		int tcphlen;
239 
240 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
241 		    n->m_next == NULL)
242 			goto stdreply;
243 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
244 		    (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
245 			goto freeit;
246 		oip = mtod(n, struct ip *);
247 		th = mtodo(n, oiphlen);
248 		tcphlen = th->th_off << 2;
249 		if (tcphlen < sizeof(struct tcphdr))
250 			goto freeit;
251 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
252 			goto freeit;
253 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
254 			goto stdreply;
255 		if (n->m_len < oiphlen + tcphlen &&
256 		    (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
257 			goto freeit;
258 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
259 		    ntohs(oip->ip_len) - oiphlen));
260 	} else if (oip->ip_p == IPPROTO_SCTP) {
261 		struct sctphdr *sh;
262 		struct sctp_chunkhdr *ch;
263 
264 		if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
265 			goto stdreply;
266 		if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
267 		    n->m_next == NULL)
268 			goto stdreply;
269 		if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
270 		    (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
271 			goto freeit;
272 		oip = mtod(n, struct ip *);
273 		icmpelen = max(sizeof(struct sctphdr),
274 		    min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
275 		sh = mtodo(n, oiphlen);
276 		if (ntohl(sh->v_tag) == 0 &&
277 		    ntohs(oip->ip_len) >= oiphlen +
278 		    sizeof(struct sctphdr) + 8 &&
279 		    (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
280 		     n->m_next != NULL)) {
281 			if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
282 			    (n = m_pullup(n, oiphlen +
283 			    sizeof(struct sctphdr) + 8)) == NULL)
284 				goto freeit;
285 			oip = mtod(n, struct ip *);
286 			sh = mtodo(n, oiphlen);
287 			ch = (struct sctp_chunkhdr *)(sh + 1);
288 			if (ch->chunk_type == SCTP_INITIATION) {
289 				icmpelen = max(sizeof(struct sctphdr) + 8,
290 				    min(V_icmp_quotelen, ntohs(oip->ip_len) -
291 				    oiphlen));
292 			}
293 		}
294 	} else
295 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
296 		    oiphlen));
297 
298 	icmplen = min(oiphlen + icmpelen, nlen);
299 	if (icmplen < sizeof(struct ip))
300 		goto freeit;
301 
302 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
303 		m = m_gethdr(M_NOWAIT, MT_DATA);
304 	else
305 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
306 	if (m == NULL)
307 		goto freeit;
308 #ifdef MAC
309 	mac_netinet_icmp_reply(n, m);
310 #endif
311 	icmplen = min(icmplen, M_TRAILINGSPACE(m) -
312 	    sizeof(struct ip) - ICMP_MINLEN);
313 	m_align(m, ICMP_MINLEN + icmplen);
314 	m->m_len = ICMP_MINLEN + icmplen;
315 
316 	/* XXX MRT  make the outgoing packet use the same FIB
317 	 * that was associated with the incoming packet
318 	 */
319 	M_SETFIB(m, M_GETFIB(n));
320 	icp = mtod(m, struct icmp *);
321 	ICMPSTAT_INC(icps_outhist[type]);
322 	icp->icmp_type = type;
323 	if (type == ICMP_REDIRECT)
324 		icp->icmp_gwaddr.s_addr = dest;
325 	else {
326 		icp->icmp_void = 0;
327 		/*
328 		 * The following assignments assume an overlay with the
329 		 * just zeroed icmp_void field.
330 		 */
331 		if (type == ICMP_PARAMPROB) {
332 			icp->icmp_pptr = code;
333 			code = 0;
334 		} else if (type == ICMP_UNREACH &&
335 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
336 			icp->icmp_nextmtu = htons(mtu);
337 		}
338 	}
339 	icp->icmp_code = code;
340 
341 	/*
342 	 * Copy the quotation into ICMP message and
343 	 * convert quoted IP header back to network representation.
344 	 */
345 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
346 	nip = &icp->icmp_ip;
347 
348 	/*
349 	 * Set up ICMP message mbuf and copy old IP header (without options
350 	 * in front of ICMP message.
351 	 * If the original mbuf was meant to bypass the firewall, the error
352 	 * reply should bypass as well.
353 	 */
354 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
355 	m->m_data -= sizeof(struct ip);
356 	m->m_len += sizeof(struct ip);
357 	m->m_pkthdr.len = m->m_len;
358 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
359 	nip = mtod(m, struct ip *);
360 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
361 	nip->ip_len = htons(m->m_len);
362 	nip->ip_v = IPVERSION;
363 	nip->ip_hl = 5;
364 	nip->ip_p = IPPROTO_ICMP;
365 	nip->ip_tos = 0;
366 	nip->ip_off = 0;
367 	icmp_reflect(m);
368 
369 freeit:
370 	m_freem(n);
371 }
372 
373 /*
374  * Process a received ICMP message.
375  */
376 int
377 icmp_input(struct mbuf **mp, int *offp, int proto)
378 {
379 	struct icmp *icp;
380 	struct in_ifaddr *ia;
381 	struct mbuf *m = *mp;
382 	struct ip *ip = mtod(m, struct ip *);
383 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
384 	int hlen = *offp;
385 	int icmplen = ntohs(ip->ip_len) - *offp;
386 	int i, code;
387 	void (*ctlfunc)(int, struct sockaddr *, void *);
388 	int fibnum;
389 
390 	*mp = NULL;
391 
392 	/*
393 	 * Locate icmp structure in mbuf, and check
394 	 * that not corrupted and of at least minimum length.
395 	 */
396 #ifdef ICMPPRINTFS
397 	if (icmpprintfs) {
398 		char srcbuf[INET_ADDRSTRLEN];
399 		char dstbuf[INET_ADDRSTRLEN];
400 
401 		printf("icmp_input from %s to %s, len %d\n",
402 		    inet_ntoa_r(ip->ip_src, srcbuf),
403 		    inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
404 	}
405 #endif
406 	if (icmplen < ICMP_MINLEN) {
407 		ICMPSTAT_INC(icps_tooshort);
408 		goto freeit;
409 	}
410 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
411 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
412 		ICMPSTAT_INC(icps_tooshort);
413 		return (IPPROTO_DONE);
414 	}
415 	ip = mtod(m, struct ip *);
416 	m->m_len -= hlen;
417 	m->m_data += hlen;
418 	icp = mtod(m, struct icmp *);
419 	if (in_cksum(m, icmplen)) {
420 		ICMPSTAT_INC(icps_checksum);
421 		goto freeit;
422 	}
423 	m->m_len += hlen;
424 	m->m_data -= hlen;
425 
426 #ifdef ICMPPRINTFS
427 	if (icmpprintfs)
428 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
429 		    icp->icmp_code);
430 #endif
431 
432 	/*
433 	 * Message type specific processing.
434 	 */
435 	if (icp->icmp_type > ICMP_MAXTYPE)
436 		goto raw;
437 
438 	/* Initialize */
439 	bzero(&icmpsrc, sizeof(icmpsrc));
440 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
441 	icmpsrc.sin_family = AF_INET;
442 	bzero(&icmpdst, sizeof(icmpdst));
443 	icmpdst.sin_len = sizeof(struct sockaddr_in);
444 	icmpdst.sin_family = AF_INET;
445 	bzero(&icmpgw, sizeof(icmpgw));
446 	icmpgw.sin_len = sizeof(struct sockaddr_in);
447 	icmpgw.sin_family = AF_INET;
448 
449 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
450 	code = icp->icmp_code;
451 	switch (icp->icmp_type) {
452 
453 	case ICMP_UNREACH:
454 		switch (code) {
455 			case ICMP_UNREACH_NET:
456 			case ICMP_UNREACH_HOST:
457 			case ICMP_UNREACH_SRCFAIL:
458 			case ICMP_UNREACH_NET_UNKNOWN:
459 			case ICMP_UNREACH_HOST_UNKNOWN:
460 			case ICMP_UNREACH_ISOLATED:
461 			case ICMP_UNREACH_TOSNET:
462 			case ICMP_UNREACH_TOSHOST:
463 			case ICMP_UNREACH_HOST_PRECEDENCE:
464 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
465 				code = PRC_UNREACH_NET;
466 				break;
467 
468 			case ICMP_UNREACH_NEEDFRAG:
469 				code = PRC_MSGSIZE;
470 				break;
471 
472 			/*
473 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
474 			 * Treat subcodes 2,3 as immediate RST
475 			 */
476 			case ICMP_UNREACH_PROTOCOL:
477 				code = PRC_UNREACH_PROTOCOL;
478 				break;
479 			case ICMP_UNREACH_PORT:
480 				code = PRC_UNREACH_PORT;
481 				break;
482 
483 			case ICMP_UNREACH_NET_PROHIB:
484 			case ICMP_UNREACH_HOST_PROHIB:
485 			case ICMP_UNREACH_FILTER_PROHIB:
486 				code = PRC_UNREACH_ADMIN_PROHIB;
487 				break;
488 
489 			default:
490 				goto badcode;
491 		}
492 		goto deliver;
493 
494 	case ICMP_TIMXCEED:
495 		if (code > 1)
496 			goto badcode;
497 		code += PRC_TIMXCEED_INTRANS;
498 		goto deliver;
499 
500 	case ICMP_PARAMPROB:
501 		if (code > 1)
502 			goto badcode;
503 		code = PRC_PARAMPROB;
504 	deliver:
505 		/*
506 		 * Problem with datagram; advise higher level routines.
507 		 */
508 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
509 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
510 			ICMPSTAT_INC(icps_badlen);
511 			goto freeit;
512 		}
513 		/* Discard ICMP's in response to multicast packets */
514 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
515 			goto badcode;
516 #ifdef ICMPPRINTFS
517 		if (icmpprintfs)
518 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
519 #endif
520 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
521 		/*
522 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
523 		 * notification to TCP layer.
524 		 */
525 		i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
526 		ip_stripoptions(m);
527 		if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
528 			/* This should actually not happen */
529 			ICMPSTAT_INC(icps_tooshort);
530 			return (IPPROTO_DONE);
531 		}
532 		ip = mtod(m, struct ip *);
533 		icp = (struct icmp *)(ip + 1);
534 		/*
535 		 * The upper layer handler can rely on:
536 		 * - The outer IP header has no options.
537 		 * - The outer IP header, the ICMP header, the inner IP header,
538 		 *   and the first n bytes of the inner payload are contiguous.
539 		 *   n is at least 8, but might be larger based on
540 		 *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
541 		 */
542 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
543 		if (ctlfunc)
544 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
545 				   (void *)&icp->icmp_ip);
546 		break;
547 
548 	badcode:
549 		ICMPSTAT_INC(icps_badcode);
550 		break;
551 
552 	case ICMP_ECHO:
553 		if (!V_icmpbmcastecho
554 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
555 			ICMPSTAT_INC(icps_bmcastecho);
556 			break;
557 		}
558 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
559 			goto freeit;
560 		icp->icmp_type = ICMP_ECHOREPLY;
561 		goto reflect;
562 
563 	case ICMP_TSTAMP:
564 		if (V_icmptstamprepl == 0)
565 			break;
566 		if (!V_icmpbmcastecho
567 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
568 			ICMPSTAT_INC(icps_bmcasttstamp);
569 			break;
570 		}
571 		if (icmplen < ICMP_TSLEN) {
572 			ICMPSTAT_INC(icps_badlen);
573 			break;
574 		}
575 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
576 			goto freeit;
577 		icp->icmp_type = ICMP_TSTAMPREPLY;
578 		icp->icmp_rtime = iptime();
579 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
580 		goto reflect;
581 
582 	case ICMP_MASKREQ:
583 		if (V_icmpmaskrepl == 0)
584 			break;
585 		/*
586 		 * We are not able to respond with all ones broadcast
587 		 * unless we receive it over a point-to-point interface.
588 		 */
589 		if (icmplen < ICMP_MASKLEN)
590 			break;
591 		switch (ip->ip_dst.s_addr) {
592 
593 		case INADDR_BROADCAST:
594 		case INADDR_ANY:
595 			icmpdst.sin_addr = ip->ip_src;
596 			break;
597 
598 		default:
599 			icmpdst.sin_addr = ip->ip_dst;
600 		}
601 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
602 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
603 		if (ia == NULL)
604 			break;
605 		if (ia->ia_ifp == NULL) {
606 			ifa_free(&ia->ia_ifa);
607 			break;
608 		}
609 		icp->icmp_type = ICMP_MASKREPLY;
610 		if (V_icmpmaskfake == 0)
611 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
612 		else
613 			icp->icmp_mask = V_icmpmaskfake;
614 		if (ip->ip_src.s_addr == 0) {
615 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
616 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
617 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
618 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
619 		}
620 		ifa_free(&ia->ia_ifa);
621 reflect:
622 		ICMPSTAT_INC(icps_reflect);
623 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
624 		icmp_reflect(m);
625 		return (IPPROTO_DONE);
626 
627 	case ICMP_REDIRECT:
628 		if (V_log_redirect) {
629 			u_long src, dst, gw;
630 
631 			src = ntohl(ip->ip_src.s_addr);
632 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
633 			gw = ntohl(icp->icmp_gwaddr.s_addr);
634 			printf("icmp redirect from %d.%d.%d.%d: "
635 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
636 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
637 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
638 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
639 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
640 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
641 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
642 		}
643 		/*
644 		 * RFC1812 says we must ignore ICMP redirects if we
645 		 * are acting as router.
646 		 */
647 		if (V_drop_redirect || V_ipforwarding)
648 			break;
649 		if (code > 3)
650 			goto badcode;
651 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
652 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
653 			ICMPSTAT_INC(icps_badlen);
654 			break;
655 		}
656 		/*
657 		 * Short circuit routing redirects to force
658 		 * immediate change in the kernel's routing
659 		 * tables.  The message is also handed to anyone
660 		 * listening on a raw socket (e.g. the routing
661 		 * daemon for use in updating its tables).
662 		 */
663 		icmpgw.sin_addr = ip->ip_src;
664 		icmpdst.sin_addr = icp->icmp_gwaddr;
665 #ifdef	ICMPPRINTFS
666 		if (icmpprintfs) {
667 			char dstbuf[INET_ADDRSTRLEN];
668 			char gwbuf[INET_ADDRSTRLEN];
669 
670 			printf("redirect dst %s to %s\n",
671 			       inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
672 			       inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
673 		}
674 #endif
675 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
676 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
677 			in_rtredirect((struct sockaddr *)&icmpsrc,
678 			  (struct sockaddr *)&icmpdst,
679 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
680 			  (struct sockaddr *)&icmpgw, fibnum);
681 		}
682 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
683 		break;
684 
685 	/*
686 	 * No kernel processing for the following;
687 	 * just fall through to send to raw listener.
688 	 */
689 	case ICMP_ECHOREPLY:
690 	case ICMP_ROUTERADVERT:
691 	case ICMP_ROUTERSOLICIT:
692 	case ICMP_TSTAMPREPLY:
693 	case ICMP_IREQREPLY:
694 	case ICMP_MASKREPLY:
695 	case ICMP_SOURCEQUENCH:
696 	default:
697 		break;
698 	}
699 
700 raw:
701 	*mp = m;
702 	rip_input(mp, offp, proto);
703 	return (IPPROTO_DONE);
704 
705 freeit:
706 	m_freem(m);
707 	return (IPPROTO_DONE);
708 }
709 
710 /*
711  * Reflect the ip packet back to the source
712  */
713 static void
714 icmp_reflect(struct mbuf *m)
715 {
716 	struct rm_priotracker in_ifa_tracker;
717 	struct ip *ip = mtod(m, struct ip *);
718 	struct ifaddr *ifa;
719 	struct ifnet *ifp;
720 	struct in_ifaddr *ia;
721 	struct in_addr t;
722 	struct nhop4_extended nh_ext;
723 	struct mbuf *opts = NULL;
724 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
725 
726 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
727 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
728 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
729 		m_freem(m);	/* Bad return address */
730 		ICMPSTAT_INC(icps_badaddr);
731 		goto done;	/* Ip_output() will check for broadcast */
732 	}
733 
734 	t = ip->ip_dst;
735 	ip->ip_dst = ip->ip_src;
736 
737 	/*
738 	 * Source selection for ICMP replies:
739 	 *
740 	 * If the incoming packet was addressed directly to one of our
741 	 * own addresses, use dst as the src for the reply.
742 	 */
743 	IN_IFADDR_RLOCK(&in_ifa_tracker);
744 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
745 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
746 			t = IA_SIN(ia)->sin_addr;
747 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
748 			goto match;
749 		}
750 	}
751 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
752 
753 	/*
754 	 * If the incoming packet was addressed to one of our broadcast
755 	 * addresses, use the first non-broadcast address which corresponds
756 	 * to the incoming interface.
757 	 */
758 	ifp = m->m_pkthdr.rcvif;
759 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
760 		IF_ADDR_RLOCK(ifp);
761 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
762 			if (ifa->ifa_addr->sa_family != AF_INET)
763 				continue;
764 			ia = ifatoia(ifa);
765 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
766 			    t.s_addr) {
767 				t = IA_SIN(ia)->sin_addr;
768 				IF_ADDR_RUNLOCK(ifp);
769 				goto match;
770 			}
771 		}
772 		IF_ADDR_RUNLOCK(ifp);
773 	}
774 	/*
775 	 * If the packet was transiting through us, use the address of
776 	 * the interface the packet came through in.  If that interface
777 	 * doesn't have a suitable IP address, the normal selection
778 	 * criteria apply.
779 	 */
780 	if (V_icmp_rfi && ifp != NULL) {
781 		IF_ADDR_RLOCK(ifp);
782 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
783 			if (ifa->ifa_addr->sa_family != AF_INET)
784 				continue;
785 			ia = ifatoia(ifa);
786 			t = IA_SIN(ia)->sin_addr;
787 			IF_ADDR_RUNLOCK(ifp);
788 			goto match;
789 		}
790 		IF_ADDR_RUNLOCK(ifp);
791 	}
792 	/*
793 	 * If the incoming packet was not addressed directly to us, use
794 	 * designated interface for icmp replies specified by sysctl
795 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
796 	 * with normal source selection.
797 	 */
798 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
799 		IF_ADDR_RLOCK(ifp);
800 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
801 			if (ifa->ifa_addr->sa_family != AF_INET)
802 				continue;
803 			ia = ifatoia(ifa);
804 			t = IA_SIN(ia)->sin_addr;
805 			IF_ADDR_RUNLOCK(ifp);
806 			goto match;
807 		}
808 		IF_ADDR_RUNLOCK(ifp);
809 	}
810 	/*
811 	 * If the packet was transiting through us, use the address of
812 	 * the interface that is the closest to the packet source.
813 	 * When we don't have a route back to the packet source, stop here
814 	 * and drop the packet.
815 	 */
816 	if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
817 		m_freem(m);
818 		ICMPSTAT_INC(icps_noroute);
819 		goto done;
820 	}
821 	t = nh_ext.nh_src;
822 match:
823 #ifdef MAC
824 	mac_netinet_icmp_replyinplace(m);
825 #endif
826 	ip->ip_src = t;
827 	ip->ip_ttl = V_ip_defttl;
828 
829 	if (optlen > 0) {
830 		u_char *cp;
831 		int opt, cnt;
832 		u_int len;
833 
834 		/*
835 		 * Retrieve any source routing from the incoming packet;
836 		 * add on any record-route or timestamp options.
837 		 */
838 		cp = (u_char *) (ip + 1);
839 		if ((opts = ip_srcroute(m)) == NULL &&
840 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
841 			opts->m_len = sizeof(struct in_addr);
842 			mtod(opts, struct in_addr *)->s_addr = 0;
843 		}
844 		if (opts) {
845 #ifdef ICMPPRINTFS
846 		    if (icmpprintfs)
847 			    printf("icmp_reflect optlen %d rt %d => ",
848 				optlen, opts->m_len);
849 #endif
850 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
851 			    opt = cp[IPOPT_OPTVAL];
852 			    if (opt == IPOPT_EOL)
853 				    break;
854 			    if (opt == IPOPT_NOP)
855 				    len = 1;
856 			    else {
857 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
858 					    break;
859 				    len = cp[IPOPT_OLEN];
860 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
861 				        len > cnt)
862 					    break;
863 			    }
864 			    /*
865 			     * Should check for overflow, but it "can't happen"
866 			     */
867 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
868 				opt == IPOPT_SECURITY) {
869 				    bcopy((caddr_t)cp,
870 					mtod(opts, caddr_t) + opts->m_len, len);
871 				    opts->m_len += len;
872 			    }
873 		    }
874 		    /* Terminate & pad, if necessary */
875 		    cnt = opts->m_len % 4;
876 		    if (cnt) {
877 			    for (; cnt < 4; cnt++) {
878 				    *(mtod(opts, caddr_t) + opts->m_len) =
879 					IPOPT_EOL;
880 				    opts->m_len++;
881 			    }
882 		    }
883 #ifdef ICMPPRINTFS
884 		    if (icmpprintfs)
885 			    printf("%d\n", opts->m_len);
886 #endif
887 		}
888 		ip_stripoptions(m);
889 	}
890 	m_tag_delete_nonpersistent(m);
891 	m->m_flags &= ~(M_BCAST|M_MCAST);
892 	icmp_send(m, opts);
893 done:
894 	if (opts)
895 		(void)m_free(opts);
896 }
897 
898 /*
899  * Send an icmp packet back to the ip level,
900  * after supplying a checksum.
901  */
902 static void
903 icmp_send(struct mbuf *m, struct mbuf *opts)
904 {
905 	struct ip *ip = mtod(m, struct ip *);
906 	int hlen;
907 	struct icmp *icp;
908 
909 	hlen = ip->ip_hl << 2;
910 	m->m_data += hlen;
911 	m->m_len -= hlen;
912 	icp = mtod(m, struct icmp *);
913 	icp->icmp_cksum = 0;
914 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
915 	m->m_data -= hlen;
916 	m->m_len += hlen;
917 	m->m_pkthdr.rcvif = (struct ifnet *)0;
918 #ifdef ICMPPRINTFS
919 	if (icmpprintfs) {
920 		char dstbuf[INET_ADDRSTRLEN];
921 		char srcbuf[INET_ADDRSTRLEN];
922 
923 		printf("icmp_send dst %s src %s\n",
924 		    inet_ntoa_r(ip->ip_dst, dstbuf),
925 		    inet_ntoa_r(ip->ip_src, srcbuf));
926 	}
927 #endif
928 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
929 }
930 
931 /*
932  * Return milliseconds since 00:00 UTC in network format.
933  */
934 uint32_t
935 iptime(void)
936 {
937 	struct timeval atv;
938 	u_long t;
939 
940 	getmicrotime(&atv);
941 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
942 	return (htonl(t));
943 }
944 
945 /*
946  * Return the next larger or smaller MTU plateau (table from RFC 1191)
947  * given current value MTU.  If DIR is less than zero, a larger plateau
948  * is returned; otherwise, a smaller value is returned.
949  */
950 int
951 ip_next_mtu(int mtu, int dir)
952 {
953 	static int mtutab[] = {
954 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
955 		296, 68, 0
956 	};
957 	int i, size;
958 
959 	size = (sizeof mtutab) / (sizeof mtutab[0]);
960 	if (dir >= 0) {
961 		for (i = 0; i < size; i++)
962 			if (mtu > mtutab[i])
963 				return mtutab[i];
964 	} else {
965 		for (i = size - 1; i >= 0; i--)
966 			if (mtu < mtutab[i])
967 				return mtutab[i];
968 		if (mtu == mtutab[0])
969 			return mtutab[0];
970 	}
971 	return 0;
972 }
973 #endif /* INET */
974 
975 
976 /*
977  * badport_bandlim() - check for ICMP bandwidth limit
978  *
979  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
980  *	hit our bandwidth limit and it is not ok.
981  *
982  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
983  *
984  *	For now we separate the TCP and UDP subsystems w/ different 'which'
985  *	values.  We may eventually remove this separation (and simplify the
986  *	code further).
987  *
988  *	Note that the printing of the error message is delayed so we can
989  *	properly print the icmp error rate that the system was trying to do
990  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
991  *	the 'final' error, but it doesn't make sense to solve the printing
992  *	delay with more complex code.
993  */
994 struct icmp_rate {
995 	const char *descr;
996 	struct counter_rate cr;
997 };
998 static VNET_DEFINE(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
999 	{ "icmp unreach response" },
1000 	{ "icmp ping response" },
1001 	{ "icmp tstamp response" },
1002 	{ "closed port RST response" },
1003 	{ "open port RST response" },
1004 	{ "icmp6 unreach response" },
1005 	{ "sctp ootb response" }
1006 };
1007 #define	V_icmp_rates	VNET(icmp_rates)
1008 
1009 static void
1010 icmp_bandlimit_init(void)
1011 {
1012 
1013 	for (int i = 0; i < BANDLIM_MAX; i++) {
1014 		V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1015 		V_icmp_rates[i].cr.cr_ticks = ticks;
1016 	}
1017 }
1018 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1019     icmp_bandlimit_init, NULL);
1020 
1021 static void
1022 icmp_bandlimit_uninit(void)
1023 {
1024 
1025 	for (int i = 0; i < BANDLIM_MAX; i++)
1026 		counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1027 }
1028 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1029     icmp_bandlimit_uninit, NULL);
1030 
1031 int
1032 badport_bandlim(int which)
1033 {
1034 	int64_t pps;
1035 
1036 	if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1037 		return (0);
1038 
1039 	KASSERT(which >= 0 && which < BANDLIM_MAX,
1040 	    ("%s: which %d", __func__, which));
1041 
1042 	pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim);
1043 	if (pps == -1)
1044 		return (-1);
1045 	if (pps > 0 && V_icmplim_output)
1046 		log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1047 			V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim);
1048 	return (0);
1049 }
1050