xref: /freebsd-14.2/sys/netinet/ip_icmp.c (revision 2472e352)
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 #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/sysctl.h>
46 #include <sys/syslog.h>
47 
48 #include <net/if.h>
49 #include <net/if_var.h>
50 #include <net/if_private.h>
51 #include <net/if_types.h>
52 #include <net/route.h>
53 #include <net/route/route_ctl.h>
54 #include <net/route/nhop.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/tcpip.h>
69 #include <netinet/icmp_var.h>
70 
71 #ifdef INET
72 
73 #include <machine/in_cksum.h>
74 
75 #include <security/mac/mac_framework.h>
76 #endif /* INET */
77 
78 extern ipproto_ctlinput_t	*ip_ctlprotox[];
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 static int sysctl_icmplim_and_jitter(SYSCTL_HANDLER_ARGS);
86 VNET_DEFINE_STATIC(u_int, icmplim) = 200;
87 #define	V_icmplim			VNET(icmplim)
88 SYSCTL_PROC(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLTYPE_UINT |
89     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmplim), 0,
90     &sysctl_icmplim_and_jitter, "IU",
91     "Maximum number of ICMP responses per second");
92 
93 VNET_DEFINE_STATIC(int, icmplim_curr_jitter) = 0;
94 #define V_icmplim_curr_jitter		VNET(icmplim_curr_jitter)
95 VNET_DEFINE_STATIC(u_int, icmplim_jitter) = 16;
96 #define	V_icmplim_jitter		VNET(icmplim_jitter)
97 SYSCTL_PROC(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLTYPE_UINT |
98     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmplim_jitter), 0,
99     &sysctl_icmplim_and_jitter, "IU",
100     "Random icmplim jitter adjustment limit");
101 
102 VNET_DEFINE_STATIC(int, icmplim_output) = 1;
103 #define	V_icmplim_output		VNET(icmplim_output)
104 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
105 	&VNET_NAME(icmplim_output), 0,
106 	"Enable logging of ICMP response rate limiting");
107 
108 #ifdef INET
109 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
110 VNET_PCPUSTAT_SYSINIT(icmpstat);
111 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
112     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
113 
114 #ifdef VIMAGE
115 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
116 #endif /* VIMAGE */
117 
118 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
119 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
120 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
121 	&VNET_NAME(icmpmaskrepl), 0,
122 	"Reply to ICMP Address Mask Request packets");
123 
124 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
125 #define	V_icmpmaskfake			VNET(icmpmaskfake)
126 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
127 	&VNET_NAME(icmpmaskfake), 0,
128 	"Fake reply to ICMP Address Mask Request packets");
129 
130 VNET_DEFINE(int, drop_redirect) = 0;
131 #define	V_drop_redirect			VNET(drop_redirect)
132 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
133 	&VNET_NAME(drop_redirect), 0,
134 	"Ignore ICMP redirects");
135 
136 VNET_DEFINE_STATIC(int, log_redirect) = 0;
137 #define	V_log_redirect			VNET(log_redirect)
138 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
139 	&VNET_NAME(log_redirect), 0,
140 	"Log ICMP redirects to the console");
141 
142 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */
143 #define	V_redirtimeout			VNET(redirtimeout)
144 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW,
145 	&VNET_NAME(redirtimeout), 0,
146 	"Delay in seconds before expiring redirect route");
147 
148 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
149 #define	V_reply_src			VNET(reply_src)
150 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
151 	&VNET_NAME(reply_src), IFNAMSIZ,
152 	"ICMP reply source for non-local packets");
153 
154 VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
155 #define	V_icmp_rfi			VNET(icmp_rfi)
156 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
157 	&VNET_NAME(icmp_rfi), 0,
158 	"ICMP reply from incoming interface for non-local packets");
159 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
160 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
161 #define	V_icmp_quotelen			VNET(icmp_quotelen)
162 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
163 	&VNET_NAME(icmp_quotelen), 0,
164 	"Number of bytes from original packet to quote in ICMP reply");
165 
166 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
167 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
168 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
169 	&VNET_NAME(icmpbmcastecho), 0,
170 	"Reply to multicast ICMP Echo Request and Timestamp packets");
171 
172 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
173 #define	V_icmptstamprepl		VNET(icmptstamprepl)
174 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
175 	&VNET_NAME(icmptstamprepl), 0,
176 	"Respond to ICMP Timestamp packets");
177 
178 VNET_DEFINE_STATIC(int, error_keeptags) = 0;
179 #define	V_error_keeptags		VNET(error_keeptags)
180 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
181 	&VNET_NAME(error_keeptags), 0,
182 	"ICMP error response keeps copy of mbuf_tags of original packet");
183 
184 #ifdef ICMPPRINTFS
185 int	icmpprintfs = 0;
186 #endif
187 
188 static void	icmp_reflect(struct mbuf *);
189 static void	icmp_send(struct mbuf *, struct mbuf *);
190 static int	icmp_verify_redirect_gateway(struct sockaddr_in *,
191     struct sockaddr_in *, struct sockaddr_in *, u_int);
192 
193 /*
194  * Kernel module interface for updating icmpstat.  The argument is an index
195  * into icmpstat treated as an array of u_long.  While this encodes the
196  * general layout of icmpstat into the caller, it doesn't encode its
197  * location, so that future changes to add, for example, per-CPU stats
198  * support won't cause binary compatibility problems for kernel modules.
199  */
200 void
kmod_icmpstat_inc(int statnum)201 kmod_icmpstat_inc(int statnum)
202 {
203 
204 	counter_u64_add(VNET(icmpstat)[statnum], 1);
205 }
206 
207 /*
208  * Generate an error packet of type error
209  * in response to bad packet ip.
210  */
211 void
icmp_error(struct mbuf * n,int type,int code,uint32_t dest,int mtu)212 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
213 {
214 	struct ip *oip, *nip;
215 	struct icmp *icp;
216 	struct mbuf *m;
217 	unsigned icmplen, icmpelen, nlen, oiphlen;
218 
219 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
220 	    __func__));
221 
222 	if (type != ICMP_REDIRECT)
223 		ICMPSTAT_INC(icps_error);
224 	/*
225 	 * Don't send error:
226 	 *  if the original packet was encrypted.
227 	 *  if not the first fragment of message.
228 	 *  in response to a multicast or broadcast packet.
229 	 *  if the old packet protocol was an ICMP error message.
230 	 */
231 	if (n->m_flags & M_DECRYPTED)
232 		goto freeit;
233 	if (n->m_flags & (M_BCAST|M_MCAST))
234 		goto freeit;
235 
236 	/* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
237 	if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
238 		goto freeit;
239 	oip = mtod(n, struct ip *);
240 	oiphlen = oip->ip_hl << 2;
241 	if (n->m_len < oiphlen + ICMP_MINLEN)
242 		goto freeit;
243 #ifdef ICMPPRINTFS
244 	if (icmpprintfs)
245 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
246 #endif
247 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
248 		goto freeit;
249 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
250 	    !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
251 		oiphlen))->icmp_type)) {
252 		ICMPSTAT_INC(icps_oldicmp);
253 		goto freeit;
254 	}
255 	/*
256 	 * Calculate length to quote from original packet and
257 	 * prevent the ICMP mbuf from overflowing.
258 	 * Unfortunately this is non-trivial since ip_forward()
259 	 * sends us truncated packets.
260 	 */
261 	nlen = m_length(n, NULL);
262 	if (oip->ip_p == IPPROTO_TCP) {
263 		struct tcphdr *th;
264 		int tcphlen;
265 
266 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
267 		    n->m_next == NULL)
268 			goto stdreply;
269 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
270 		    (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
271 			goto freeit;
272 		oip = mtod(n, struct ip *);
273 		th = mtodo(n, oiphlen);
274 		tcphlen = th->th_off << 2;
275 		if (tcphlen < sizeof(struct tcphdr))
276 			goto freeit;
277 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
278 			goto freeit;
279 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
280 			goto stdreply;
281 		if (n->m_len < oiphlen + tcphlen &&
282 		    (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
283 			goto freeit;
284 		oip = mtod(n, struct ip *);
285 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
286 		    ntohs(oip->ip_len) - oiphlen));
287 	} else if (oip->ip_p == IPPROTO_SCTP) {
288 		struct sctphdr *sh;
289 		struct sctp_chunkhdr *ch;
290 
291 		if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
292 			goto stdreply;
293 		if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
294 		    n->m_next == NULL)
295 			goto stdreply;
296 		if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
297 		    (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
298 			goto freeit;
299 		oip = mtod(n, struct ip *);
300 		icmpelen = max(sizeof(struct sctphdr),
301 		    min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
302 		sh = mtodo(n, oiphlen);
303 		if (ntohl(sh->v_tag) == 0 &&
304 		    ntohs(oip->ip_len) >= oiphlen +
305 		    sizeof(struct sctphdr) + 8 &&
306 		    (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
307 		     n->m_next != NULL)) {
308 			if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
309 			    (n = m_pullup(n, oiphlen +
310 			    sizeof(struct sctphdr) + 8)) == NULL)
311 				goto freeit;
312 			oip = mtod(n, struct ip *);
313 			sh = mtodo(n, oiphlen);
314 			ch = (struct sctp_chunkhdr *)(sh + 1);
315 			if (ch->chunk_type == SCTP_INITIATION) {
316 				icmpelen = max(sizeof(struct sctphdr) + 8,
317 				    min(V_icmp_quotelen, ntohs(oip->ip_len) -
318 				    oiphlen));
319 			}
320 		}
321 	} else
322 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
323 		    oiphlen));
324 
325 	icmplen = min(oiphlen + icmpelen, nlen);
326 	if (icmplen < sizeof(struct ip))
327 		goto freeit;
328 
329 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
330 		m = m_gethdr(M_NOWAIT, MT_DATA);
331 	else
332 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
333 	if (m == NULL)
334 		goto freeit;
335 #ifdef MAC
336 	mac_netinet_icmp_reply(n, m);
337 #endif
338 	icmplen = min(icmplen, M_TRAILINGSPACE(m) -
339 	    sizeof(struct ip) - ICMP_MINLEN);
340 	m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
341 	m->m_data += sizeof(struct ip);
342 	m->m_len = ICMP_MINLEN + icmplen;
343 
344 	/* XXX MRT  make the outgoing packet use the same FIB
345 	 * that was associated with the incoming packet
346 	 */
347 	M_SETFIB(m, M_GETFIB(n));
348 	icp = mtod(m, struct icmp *);
349 	ICMPSTAT_INC(icps_outhist[type]);
350 	icp->icmp_type = type;
351 	if (type == ICMP_REDIRECT)
352 		icp->icmp_gwaddr.s_addr = dest;
353 	else {
354 		icp->icmp_void = 0;
355 		/*
356 		 * The following assignments assume an overlay with the
357 		 * just zeroed icmp_void field.
358 		 */
359 		if (type == ICMP_PARAMPROB) {
360 			icp->icmp_pptr = code;
361 			code = 0;
362 		} else if (type == ICMP_UNREACH &&
363 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
364 			icp->icmp_nextmtu = htons(mtu);
365 		}
366 	}
367 	icp->icmp_code = code;
368 
369 	/*
370 	 * Copy the quotation into ICMP message and
371 	 * convert quoted IP header back to network representation.
372 	 */
373 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
374 	nip = &icp->icmp_ip;
375 
376 	/*
377 	 * Set up ICMP message mbuf and copy old IP header (without options
378 	 * in front of ICMP message.
379 	 * If the original mbuf was meant to bypass the firewall, the error
380 	 * reply should bypass as well.
381 	 */
382 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
383 	KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
384 	    ("insufficient space for ip header"));
385 	m->m_data -= sizeof(struct ip);
386 	m->m_len += sizeof(struct ip);
387 	m->m_pkthdr.len = m->m_len;
388 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
389 	nip = mtod(m, struct ip *);
390 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
391 	nip->ip_len = htons(m->m_len);
392 	nip->ip_v = IPVERSION;
393 	nip->ip_hl = 5;
394 	nip->ip_p = IPPROTO_ICMP;
395 	nip->ip_tos = 0;
396 	nip->ip_off = 0;
397 
398 	if (V_error_keeptags)
399 		m_tag_copy_chain(m, n, M_NOWAIT);
400 
401 	icmp_reflect(m);
402 
403 freeit:
404 	m_freem(n);
405 }
406 
407 int
icmp_errmap(const struct icmp * icp)408 icmp_errmap(const struct icmp *icp)
409 {
410 
411 	switch (icp->icmp_type) {
412 	case ICMP_UNREACH:
413 		switch (icp->icmp_code) {
414 		case ICMP_UNREACH_NET:
415 		case ICMP_UNREACH_HOST:
416 		case ICMP_UNREACH_SRCFAIL:
417 		case ICMP_UNREACH_NET_UNKNOWN:
418 		case ICMP_UNREACH_HOST_UNKNOWN:
419 		case ICMP_UNREACH_ISOLATED:
420 		case ICMP_UNREACH_TOSNET:
421 		case ICMP_UNREACH_TOSHOST:
422 		case ICMP_UNREACH_HOST_PRECEDENCE:
423 		case ICMP_UNREACH_PRECEDENCE_CUTOFF:
424 			return (EHOSTUNREACH);
425 		case ICMP_UNREACH_NEEDFRAG:
426 			return (EMSGSIZE);
427 		case ICMP_UNREACH_PROTOCOL:
428 		case ICMP_UNREACH_PORT:
429 		case ICMP_UNREACH_NET_PROHIB:
430 		case ICMP_UNREACH_HOST_PROHIB:
431 		case ICMP_UNREACH_FILTER_PROHIB:
432 			return (ECONNREFUSED);
433 		default:
434 			return (0);
435 		}
436 	case ICMP_TIMXCEED:
437 		switch (icp->icmp_code) {
438 		case ICMP_TIMXCEED_INTRANS:
439 			return (EHOSTUNREACH);
440 		default:
441 			return (0);
442 		}
443 	case ICMP_PARAMPROB:
444 		switch (icp->icmp_code) {
445 		case ICMP_PARAMPROB_ERRATPTR:
446 		case ICMP_PARAMPROB_OPTABSENT:
447 			return (ENOPROTOOPT);
448 		default:
449 			return (0);
450 		}
451 	default:
452 		return (0);
453 	}
454 }
455 
456 /*
457  * Process a received ICMP message.
458  */
459 int
icmp_input(struct mbuf ** mp,int * offp,int proto)460 icmp_input(struct mbuf **mp, int *offp, int proto)
461 {
462 	struct icmp *icp;
463 	struct in_ifaddr *ia;
464 	struct mbuf *m = *mp;
465 	struct ip *ip = mtod(m, struct ip *);
466 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
467 	int hlen = *offp;
468 	int icmplen = ntohs(ip->ip_len) - *offp;
469 	int i, code;
470 	int fibnum;
471 
472 	NET_EPOCH_ASSERT();
473 
474 	*mp = NULL;
475 
476 	/*
477 	 * Locate icmp structure in mbuf, and check
478 	 * that not corrupted and of at least minimum length.
479 	 */
480 #ifdef ICMPPRINTFS
481 	if (icmpprintfs) {
482 		char srcbuf[INET_ADDRSTRLEN];
483 		char dstbuf[INET_ADDRSTRLEN];
484 
485 		printf("icmp_input from %s to %s, len %d\n",
486 		    inet_ntoa_r(ip->ip_src, srcbuf),
487 		    inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
488 	}
489 #endif
490 	if (icmplen < ICMP_MINLEN) {
491 		ICMPSTAT_INC(icps_tooshort);
492 		goto freeit;
493 	}
494 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
495 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
496 		ICMPSTAT_INC(icps_tooshort);
497 		return (IPPROTO_DONE);
498 	}
499 	ip = mtod(m, struct ip *);
500 	m->m_len -= hlen;
501 	m->m_data += hlen;
502 	icp = mtod(m, struct icmp *);
503 	if (in_cksum(m, icmplen)) {
504 		ICMPSTAT_INC(icps_checksum);
505 		goto freeit;
506 	}
507 	m->m_len += hlen;
508 	m->m_data -= hlen;
509 
510 #ifdef ICMPPRINTFS
511 	if (icmpprintfs)
512 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
513 		    icp->icmp_code);
514 #endif
515 
516 	/*
517 	 * Message type specific processing.
518 	 */
519 	if (icp->icmp_type > ICMP_MAXTYPE)
520 		goto raw;
521 
522 	/* Initialize */
523 	bzero(&icmpsrc, sizeof(icmpsrc));
524 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
525 	icmpsrc.sin_family = AF_INET;
526 	bzero(&icmpdst, sizeof(icmpdst));
527 	icmpdst.sin_len = sizeof(struct sockaddr_in);
528 	icmpdst.sin_family = AF_INET;
529 	bzero(&icmpgw, sizeof(icmpgw));
530 	icmpgw.sin_len = sizeof(struct sockaddr_in);
531 	icmpgw.sin_family = AF_INET;
532 
533 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
534 	code = icp->icmp_code;
535 	switch (icp->icmp_type) {
536 	case ICMP_UNREACH:
537 		if (code > ICMP_UNREACH_PRECEDENCE_CUTOFF)
538 			goto badcode;
539 		else
540 			goto deliver;
541 
542 	case ICMP_TIMXCEED:
543 		if (code > ICMP_TIMXCEED_REASS)
544 			goto badcode;
545 		else
546 			goto deliver;
547 
548 	case ICMP_PARAMPROB:
549 		if (code > ICMP_PARAMPROB_LENGTH)
550 			goto badcode;
551 
552 	deliver:
553 		/*
554 		 * Problem with datagram; advise higher level routines.
555 		 */
556 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
557 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
558 			ICMPSTAT_INC(icps_badlen);
559 			goto freeit;
560 		}
561 		/* Discard ICMP's in response to multicast packets */
562 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
563 			goto badcode;
564 		/* Filter out responses to INADDR_ANY, protocols ignore it. */
565 		if (icp->icmp_ip.ip_dst.s_addr == INADDR_ANY ||
566 		    icp->icmp_ip.ip_src.s_addr == INADDR_ANY)
567 			goto freeit;
568 #ifdef ICMPPRINTFS
569 		if (icmpprintfs)
570 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
571 #endif
572 		/*
573 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
574 		 * notification to TCP layer.
575 		 */
576 		i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
577 		ip_stripoptions(m);
578 		if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
579 			/* This should actually not happen */
580 			ICMPSTAT_INC(icps_tooshort);
581 			return (IPPROTO_DONE);
582 		}
583 		ip = mtod(m, struct ip *);
584 		icp = (struct icmp *)(ip + 1);
585 		/*
586 		 * The upper layer handler can rely on:
587 		 * - The outer IP header has no options.
588 		 * - The outer IP header, the ICMP header, the inner IP header,
589 		 *   and the first n bytes of the inner payload are contiguous.
590 		 *   n is at least 8, but might be larger based on
591 		 *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
592 		 */
593 		if (ip_ctlprotox[icp->icmp_ip.ip_p] != NULL)
594 			ip_ctlprotox[icp->icmp_ip.ip_p](icp);
595 		break;
596 
597 	badcode:
598 		ICMPSTAT_INC(icps_badcode);
599 		break;
600 
601 	case ICMP_ECHO:
602 		if (!V_icmpbmcastecho
603 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
604 			ICMPSTAT_INC(icps_bmcastecho);
605 			break;
606 		}
607 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
608 			goto freeit;
609 		icp->icmp_type = ICMP_ECHOREPLY;
610 		goto reflect;
611 
612 	case ICMP_TSTAMP:
613 		if (V_icmptstamprepl == 0)
614 			break;
615 		if (!V_icmpbmcastecho
616 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
617 			ICMPSTAT_INC(icps_bmcasttstamp);
618 			break;
619 		}
620 		if (icmplen < ICMP_TSLEN) {
621 			ICMPSTAT_INC(icps_badlen);
622 			break;
623 		}
624 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
625 			goto freeit;
626 		icp->icmp_type = ICMP_TSTAMPREPLY;
627 		icp->icmp_rtime = iptime();
628 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
629 		goto reflect;
630 
631 	case ICMP_MASKREQ:
632 		if (V_icmpmaskrepl == 0)
633 			break;
634 		/*
635 		 * We are not able to respond with all ones broadcast
636 		 * unless we receive it over a point-to-point interface.
637 		 */
638 		if (icmplen < ICMP_MASKLEN)
639 			break;
640 		switch (ip->ip_dst.s_addr) {
641 		case INADDR_BROADCAST:
642 		case INADDR_ANY:
643 			icmpdst.sin_addr = ip->ip_src;
644 			break;
645 
646 		default:
647 			icmpdst.sin_addr = ip->ip_dst;
648 		}
649 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
650 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
651 		if (ia == NULL)
652 			break;
653 		if (ia->ia_ifp == NULL)
654 			break;
655 		icp->icmp_type = ICMP_MASKREPLY;
656 		if (V_icmpmaskfake == 0)
657 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
658 		else
659 			icp->icmp_mask = V_icmpmaskfake;
660 		if (ip->ip_src.s_addr == 0) {
661 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
662 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
663 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
664 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
665 		}
666 reflect:
667 		ICMPSTAT_INC(icps_reflect);
668 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
669 		icmp_reflect(m);
670 		return (IPPROTO_DONE);
671 
672 	case ICMP_REDIRECT:
673 		if (V_log_redirect) {
674 			u_long src, dst, gw;
675 
676 			src = ntohl(ip->ip_src.s_addr);
677 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
678 			gw = ntohl(icp->icmp_gwaddr.s_addr);
679 			printf("icmp redirect from %d.%d.%d.%d: "
680 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
681 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
682 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
683 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
684 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
685 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
686 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
687 		}
688 		/*
689 		 * RFC1812 says we must ignore ICMP redirects if we
690 		 * are acting as router.
691 		 */
692 		if (V_drop_redirect || V_ipforwarding)
693 			break;
694 		if (code > 3)
695 			goto badcode;
696 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
697 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
698 			ICMPSTAT_INC(icps_badlen);
699 			break;
700 		}
701 		/*
702 		 * Short circuit routing redirects to force
703 		 * immediate change in the kernel's routing
704 		 * tables.  The message is also handed to anyone
705 		 * listening on a raw socket (e.g. the routing
706 		 * daemon for use in updating its tables).
707 		 */
708 		icmpgw.sin_addr = ip->ip_src;
709 		icmpdst.sin_addr = icp->icmp_gwaddr;
710 #ifdef	ICMPPRINTFS
711 		if (icmpprintfs) {
712 			char dstbuf[INET_ADDRSTRLEN];
713 			char gwbuf[INET_ADDRSTRLEN];
714 
715 			printf("redirect dst %s to %s\n",
716 			       inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
717 			       inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
718 		}
719 #endif
720 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
721 
722 		/*
723 		 * RFC 1122 says network (code 0,2) redirects SHOULD
724 		 * be treated identically to the host redirects.
725 		 * Given that, ignore network masks.
726 		 */
727 
728 		/*
729 		 * Variable values:
730 		 * icmpsrc: route destination
731 		 * icmpdst: route gateway
732 		 * icmpgw: message source
733 		 */
734 
735 		if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst,
736 		    M_GETFIB(m)) != 0) {
737 			/* TODO: increment bad redirects here */
738 			break;
739 		}
740 
741 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
742 			rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc,
743 			    (struct sockaddr *)&icmpdst,
744 			    (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif,
745 			    RTF_GATEWAY, V_redirtimeout);
746 		}
747 		break;
748 
749 	/*
750 	 * No kernel processing for the following;
751 	 * just fall through to send to raw listener.
752 	 */
753 	case ICMP_ECHOREPLY:
754 	case ICMP_ROUTERADVERT:
755 	case ICMP_ROUTERSOLICIT:
756 	case ICMP_TSTAMPREPLY:
757 	case ICMP_IREQREPLY:
758 	case ICMP_MASKREPLY:
759 	case ICMP_SOURCEQUENCH:
760 	default:
761 		break;
762 	}
763 
764 raw:
765 	*mp = m;
766 	rip_input(mp, offp, proto);
767 	return (IPPROTO_DONE);
768 
769 freeit:
770 	m_freem(m);
771 	return (IPPROTO_DONE);
772 }
773 
774 /*
775  * Reflect the ip packet back to the source
776  */
777 static void
icmp_reflect(struct mbuf * m)778 icmp_reflect(struct mbuf *m)
779 {
780 	struct ip *ip = mtod(m, struct ip *);
781 	struct ifaddr *ifa;
782 	struct ifnet *ifp;
783 	struct in_ifaddr *ia;
784 	struct in_addr t;
785 	struct nhop_object *nh;
786 	struct mbuf *opts = NULL;
787 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
788 
789 	NET_EPOCH_ASSERT();
790 
791 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
792 	    (IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net240) ||
793 	    (IN_ZERONET(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net0) ) {
794 		m_freem(m);	/* Bad return address */
795 		ICMPSTAT_INC(icps_badaddr);
796 		goto done;	/* Ip_output() will check for broadcast */
797 	}
798 
799 	t = ip->ip_dst;
800 	ip->ip_dst = ip->ip_src;
801 
802 	/*
803 	 * Source selection for ICMP replies:
804 	 *
805 	 * If the incoming packet was addressed directly to one of our
806 	 * own addresses, use dst as the src for the reply.
807 	 */
808 	CK_LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
809 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
810 			t = IA_SIN(ia)->sin_addr;
811 			goto match;
812 		}
813 	}
814 
815 	/*
816 	 * If the incoming packet was addressed to one of our broadcast
817 	 * addresses, use the first non-broadcast address which corresponds
818 	 * to the incoming interface.
819 	 */
820 	ifp = m->m_pkthdr.rcvif;
821 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
822 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
823 			if (ifa->ifa_addr->sa_family != AF_INET)
824 				continue;
825 			ia = ifatoia(ifa);
826 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
827 			    t.s_addr) {
828 				t = IA_SIN(ia)->sin_addr;
829 				goto match;
830 			}
831 		}
832 	}
833 	/*
834 	 * If the packet was transiting through us, use the address of
835 	 * the interface the packet came through in.  If that interface
836 	 * doesn't have a suitable IP address, the normal selection
837 	 * criteria apply.
838 	 */
839 	if (V_icmp_rfi && ifp != NULL) {
840 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
841 			if (ifa->ifa_addr->sa_family != AF_INET)
842 				continue;
843 			ia = ifatoia(ifa);
844 			t = IA_SIN(ia)->sin_addr;
845 			goto match;
846 		}
847 	}
848 	/*
849 	 * If the incoming packet was not addressed directly to us, use
850 	 * designated interface for icmp replies specified by sysctl
851 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
852 	 * with normal source selection.
853 	 */
854 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
855 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
856 			if (ifa->ifa_addr->sa_family != AF_INET)
857 				continue;
858 			ia = ifatoia(ifa);
859 			t = IA_SIN(ia)->sin_addr;
860 			goto match;
861 		}
862 	}
863 	/*
864 	 * If the packet was transiting through us, use the address of
865 	 * the interface that is the closest to the packet source.
866 	 * When we don't have a route back to the packet source, stop here
867 	 * and drop the packet.
868 	 */
869 	nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
870 	if (nh == NULL) {
871 		m_freem(m);
872 		ICMPSTAT_INC(icps_noroute);
873 		goto done;
874 	}
875 	t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr;
876 match:
877 #ifdef MAC
878 	mac_netinet_icmp_replyinplace(m);
879 #endif
880 	ip->ip_src = t;
881 	ip->ip_ttl = V_ip_defttl;
882 
883 	if (optlen > 0) {
884 		u_char *cp;
885 		int opt, cnt;
886 		u_int len;
887 
888 		/*
889 		 * Retrieve any source routing from the incoming packet;
890 		 * add on any record-route or timestamp options.
891 		 */
892 		cp = (u_char *) (ip + 1);
893 		if ((opts = ip_srcroute(m)) == NULL &&
894 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
895 			opts->m_len = sizeof(struct in_addr);
896 			mtod(opts, struct in_addr *)->s_addr = 0;
897 		}
898 		if (opts) {
899 #ifdef ICMPPRINTFS
900 		    if (icmpprintfs)
901 			    printf("icmp_reflect optlen %d rt %d => ",
902 				optlen, opts->m_len);
903 #endif
904 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
905 			    opt = cp[IPOPT_OPTVAL];
906 			    if (opt == IPOPT_EOL)
907 				    break;
908 			    if (opt == IPOPT_NOP)
909 				    len = 1;
910 			    else {
911 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
912 					    break;
913 				    len = cp[IPOPT_OLEN];
914 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
915 				        len > cnt)
916 					    break;
917 			    }
918 			    /*
919 			     * Should check for overflow, but it "can't happen"
920 			     */
921 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
922 				opt == IPOPT_SECURITY) {
923 				    bcopy((caddr_t)cp,
924 					mtod(opts, caddr_t) + opts->m_len, len);
925 				    opts->m_len += len;
926 			    }
927 		    }
928 		    /* Terminate & pad, if necessary */
929 		    cnt = opts->m_len % 4;
930 		    if (cnt) {
931 			    for (; cnt < 4; cnt++) {
932 				    *(mtod(opts, caddr_t) + opts->m_len) =
933 					IPOPT_EOL;
934 				    opts->m_len++;
935 			    }
936 		    }
937 #ifdef ICMPPRINTFS
938 		    if (icmpprintfs)
939 			    printf("%d\n", opts->m_len);
940 #endif
941 		}
942 		ip_stripoptions(m);
943 	}
944 	m_tag_delete_nonpersistent(m);
945 	m->m_flags &= ~(M_BCAST|M_MCAST);
946 	icmp_send(m, opts);
947 done:
948 	if (opts)
949 		(void)m_free(opts);
950 }
951 
952 /*
953  * Verifies if redirect message is valid, according to RFC 1122
954  *
955  * @src: sockaddr with address of redirect originator
956  * @dst: sockaddr with destination in question
957  * @gateway: new proposed gateway
958  *
959  * Returns 0 on success.
960  */
961 static int
icmp_verify_redirect_gateway(struct sockaddr_in * src,struct sockaddr_in * dst,struct sockaddr_in * gateway,u_int fibnum)962 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst,
963     struct sockaddr_in *gateway, u_int fibnum)
964 {
965 	struct nhop_object *nh;
966 	struct ifaddr *ifa;
967 
968 	NET_EPOCH_ASSERT();
969 
970 	/* Verify the gateway is directly reachable. */
971 	if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL)
972 		return (ENETUNREACH);
973 
974 	/* TODO: fib-aware. */
975 	if (ifa_ifwithaddr_check((struct sockaddr *)gateway))
976 		return (EHOSTUNREACH);
977 
978 	nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0);
979 	if (nh == NULL)
980 		return (EINVAL);
981 
982 	/*
983 	 * If the redirect isn't from our current router for this dst,
984 	 * it's either old or wrong.  If it redirects us to ourselves,
985 	 * we have a routing loop, perhaps as a result of an interface
986 	 * going down recently.
987 	 */
988 	if (!sa_equal((struct sockaddr *)src, &nh->gw_sa))
989 		return (EINVAL);
990 	if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK)
991 		return (EINVAL);
992 
993 	/* If host route already exists, ignore redirect. */
994 	if (nh->nh_flags & NHF_HOST)
995 		return (EEXIST);
996 
997 	/* If the prefix is directly reachable, ignore redirect. */
998 	if (!(nh->nh_flags & NHF_GATEWAY))
999 		return (EEXIST);
1000 
1001 	return (0);
1002 }
1003 
1004 /*
1005  * Send an icmp packet back to the ip level,
1006  * after supplying a checksum.
1007  */
1008 static void
icmp_send(struct mbuf * m,struct mbuf * opts)1009 icmp_send(struct mbuf *m, struct mbuf *opts)
1010 {
1011 	struct ip *ip = mtod(m, struct ip *);
1012 	int hlen;
1013 	struct icmp *icp;
1014 
1015 	hlen = ip->ip_hl << 2;
1016 	m->m_data += hlen;
1017 	m->m_len -= hlen;
1018 	icp = mtod(m, struct icmp *);
1019 	icp->icmp_cksum = 0;
1020 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
1021 	m->m_data -= hlen;
1022 	m->m_len += hlen;
1023 	m->m_pkthdr.rcvif = (struct ifnet *)0;
1024 #ifdef ICMPPRINTFS
1025 	if (icmpprintfs) {
1026 		char dstbuf[INET_ADDRSTRLEN];
1027 		char srcbuf[INET_ADDRSTRLEN];
1028 
1029 		printf("icmp_send dst %s src %s\n",
1030 		    inet_ntoa_r(ip->ip_dst, dstbuf),
1031 		    inet_ntoa_r(ip->ip_src, srcbuf));
1032 	}
1033 #endif
1034 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
1035 }
1036 
1037 /*
1038  * Return milliseconds since 00:00 UTC in network format.
1039  */
1040 uint32_t
iptime(void)1041 iptime(void)
1042 {
1043 	struct timeval atv;
1044 	u_long t;
1045 
1046 	getmicrotime(&atv);
1047 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1048 	return (htonl(t));
1049 }
1050 
1051 /*
1052  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1053  * given current value MTU.  If DIR is less than zero, a larger plateau
1054  * is returned; otherwise, a smaller value is returned.
1055  */
1056 int
ip_next_mtu(int mtu,int dir)1057 ip_next_mtu(int mtu, int dir)
1058 {
1059 	static int mtutab[] = {
1060 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
1061 		296, 68, 0
1062 	};
1063 	int i, size;
1064 
1065 	size = (sizeof mtutab) / (sizeof mtutab[0]);
1066 	if (dir >= 0) {
1067 		for (i = 0; i < size; i++)
1068 			if (mtu > mtutab[i])
1069 				return mtutab[i];
1070 	} else {
1071 		for (i = size - 1; i >= 0; i--)
1072 			if (mtu < mtutab[i])
1073 				return mtutab[i];
1074 		if (mtu == mtutab[0])
1075 			return mtutab[0];
1076 	}
1077 	return 0;
1078 }
1079 #endif /* INET */
1080 
1081 /*
1082  * badport_bandlim() - check for ICMP bandwidth limit
1083  *
1084  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
1085  *	hit our bandwidth limit and it is not ok.
1086  *
1087  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
1088  *
1089  *	For now we separate the TCP and UDP subsystems w/ different 'which'
1090  *	values.  We may eventually remove this separation (and simplify the
1091  *	code further).
1092  *
1093  *	Note that the printing of the error message is delayed so we can
1094  *	properly print the icmp error rate that the system was trying to do
1095  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
1096  *	the 'final' error, but it doesn't make sense to solve the printing
1097  *	delay with more complex code.
1098  */
1099 VNET_DEFINE_STATIC(struct counter_rate, icmp_rates[BANDLIM_MAX]);
1100 #define	V_icmp_rates	VNET(icmp_rates)
1101 
1102 static const char *icmp_rate_descrs[BANDLIM_MAX] = {
1103 	[BANDLIM_ICMP_UNREACH] = "icmp unreach",
1104 	[BANDLIM_ICMP_ECHO] = "icmp ping",
1105 	[BANDLIM_ICMP_TSTAMP] = "icmp tstamp",
1106 	[BANDLIM_RST_CLOSEDPORT] = "closed port RST",
1107 	[BANDLIM_RST_OPENPORT] = "open port RST",
1108 	[BANDLIM_ICMP6_UNREACH] = "icmp6 unreach",
1109 	[BANDLIM_SCTP_OOTB] = "sctp ootb",
1110 };
1111 
1112 static void
icmplim_new_jitter(void)1113 icmplim_new_jitter(void)
1114 {
1115 	/*
1116 	 * Adjust limit +/- to jitter the measurement to deny a side-channel
1117 	 * port scan as in https://dl.acm.org/doi/10.1145/3372297.3417280
1118 	 */
1119 	if (V_icmplim_jitter > 0)
1120 		V_icmplim_curr_jitter =
1121 		    arc4random_uniform(V_icmplim_jitter * 2 + 1) -
1122 		    V_icmplim_jitter;
1123 }
1124 
1125 static int
sysctl_icmplim_and_jitter(SYSCTL_HANDLER_ARGS)1126 sysctl_icmplim_and_jitter(SYSCTL_HANDLER_ARGS)
1127 {
1128 	uint32_t new;
1129 	int error;
1130 	bool lim;
1131 
1132 	MPASS(oidp->oid_arg1 == &VNET_NAME(icmplim) ||
1133 	    oidp->oid_arg1 == &VNET_NAME(icmplim_jitter));
1134 
1135 	lim = (oidp->oid_arg1 == &VNET_NAME(icmplim));
1136 	new = lim ? V_icmplim : V_icmplim_jitter;
1137 	error = sysctl_handle_int(oidp, &new, 0, req);
1138 	if (error == 0 && req->newptr) {
1139 		if (lim) {
1140 			if (new != 0 && new <= V_icmplim_jitter)
1141 				error = EINVAL;
1142 			else
1143 				V_icmplim = new;
1144 		} else {
1145 			if (new >= V_icmplim)
1146 				error = EINVAL;
1147 			else {
1148 				V_icmplim_jitter = new;
1149 				icmplim_new_jitter();
1150 			}
1151 		}
1152 	}
1153 	MPASS(V_icmplim + V_icmplim_curr_jitter >= 0);
1154 
1155 	return (error);
1156 }
1157 
1158 static void
icmp_bandlimit_init(void)1159 icmp_bandlimit_init(void)
1160 {
1161 
1162 	for (int i = 0; i < BANDLIM_MAX; i++) {
1163 		V_icmp_rates[i].cr_rate = counter_u64_alloc(M_WAITOK);
1164 		V_icmp_rates[i].cr_ticks = ticks;
1165 	}
1166 	icmplim_new_jitter();
1167 }
1168 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1169     icmp_bandlimit_init, NULL);
1170 
1171 #ifdef VIMAGE
1172 static void
icmp_bandlimit_uninit(void)1173 icmp_bandlimit_uninit(void)
1174 {
1175 
1176 	for (int i = 0; i < BANDLIM_MAX; i++)
1177 		counter_u64_free(V_icmp_rates[i].cr_rate);
1178 }
1179 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1180     icmp_bandlimit_uninit, NULL);
1181 #endif
1182 
1183 int
badport_bandlim(int which)1184 badport_bandlim(int which)
1185 {
1186 	int64_t pps;
1187 
1188 	if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1189 		return (0);
1190 
1191 	KASSERT(which >= 0 && which < BANDLIM_MAX,
1192 	    ("%s: which %d", __func__, which));
1193 
1194 	pps = counter_ratecheck(&V_icmp_rates[which], V_icmplim +
1195 	    V_icmplim_curr_jitter);
1196 	if (pps > 0) {
1197 		if (V_icmplim_output)
1198 			log(LOG_NOTICE,
1199 			    "Limiting %s response from %jd to %d packets/sec\n",
1200 			    icmp_rate_descrs[which], (intmax_t )pps,
1201 			    V_icmplim + V_icmplim_curr_jitter);
1202 		icmplim_new_jitter();
1203 	}
1204 	if (pps == -1)
1205 		return (-1);
1206 	return (0);
1207 }
1208