xref: /freebsd-13-stable/sys/netinet/ip_icmp.c (revision e1dd07ede92382bdcc52b3093e8f2ec5d9c88467)
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/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/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/tcp_var.h>
69 #include <netinet/tcpip.h>
70 #include <netinet/icmp_var.h>
71 
72 #ifdef INET
73 
74 #include <machine/in_cksum.h>
75 
76 #include <security/mac/mac_framework.h>
77 #endif /* INET */
78 
79 /*
80  * ICMP routines: error generation, receive packet processing, and
81  * routines to turnaround packets back to the originator, and
82  * host table maintenance routines.
83  */
84 static int sysctl_icmplim_and_jitter(SYSCTL_HANDLER_ARGS);
85 VNET_DEFINE_STATIC(u_int, icmplim) = 200;
86 #define	V_icmplim			VNET(icmplim)
87 SYSCTL_PROC(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLTYPE_UINT |
88     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmplim), 0,
89     &sysctl_icmplim_and_jitter, "IU",
90     "Maximum number of ICMP responses per second");
91 
92 VNET_DEFINE_STATIC(int, icmplim_curr_jitter[BANDLIM_MAX]) = {0};
93 #define V_icmplim_curr_jitter		VNET(icmplim_curr_jitter)
94 VNET_DEFINE_STATIC(u_int, icmplim_jitter) = 16;
95 #define	V_icmplim_jitter		VNET(icmplim_jitter)
96 SYSCTL_PROC(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLTYPE_UINT |
97     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmplim_jitter), 0,
98     &sysctl_icmplim_and_jitter, "IU",
99     "Random icmplim jitter adjustment limit");
100 
101 VNET_DEFINE_STATIC(int, icmplim_output) = 1;
102 #define	V_icmplim_output		VNET(icmplim_output)
103 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
104 	&VNET_NAME(icmplim_output), 0,
105 	"Enable logging of ICMP response rate limiting");
106 
107 #ifdef INET
108 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
109 VNET_PCPUSTAT_SYSINIT(icmpstat);
110 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
111     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
112 
113 #ifdef VIMAGE
114 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
115 #endif /* VIMAGE */
116 
117 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
118 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
119 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
120 	&VNET_NAME(icmpmaskrepl), 0,
121 	"Reply to ICMP Address Mask Request packets");
122 
123 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
124 #define	V_icmpmaskfake			VNET(icmpmaskfake)
125 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
126 	&VNET_NAME(icmpmaskfake), 0,
127 	"Fake reply to ICMP Address Mask Request packets");
128 
129 VNET_DEFINE(int, drop_redirect) = 0;
130 #define	V_drop_redirect			VNET(drop_redirect)
131 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
132 	&VNET_NAME(drop_redirect), 0,
133 	"Ignore ICMP redirects");
134 
135 VNET_DEFINE_STATIC(int, log_redirect) = 0;
136 #define	V_log_redirect			VNET(log_redirect)
137 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
138 	&VNET_NAME(log_redirect), 0,
139 	"Log ICMP redirects to the console");
140 
141 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */
142 #define	V_redirtimeout			VNET(redirtimeout)
143 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW,
144 	&VNET_NAME(redirtimeout), 0,
145 	"Delay in seconds before expiring redirect route");
146 
147 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
148 #define	V_reply_src			VNET(reply_src)
149 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
150 	&VNET_NAME(reply_src), IFNAMSIZ,
151 	"ICMP reply source for non-local packets");
152 
153 VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
154 #define	V_icmp_rfi			VNET(icmp_rfi)
155 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
156 	&VNET_NAME(icmp_rfi), 0,
157 	"ICMP reply from incoming interface for non-local packets");
158 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
159 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
160 #define	V_icmp_quotelen			VNET(icmp_quotelen)
161 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
162 	&VNET_NAME(icmp_quotelen), 0,
163 	"Number of bytes from original packet to quote in ICMP reply");
164 
165 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
166 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
167 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
168 	&VNET_NAME(icmpbmcastecho), 0,
169 	"Reply to multicast ICMP Echo Request and Timestamp packets");
170 
171 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
172 #define	V_icmptstamprepl		VNET(icmptstamprepl)
173 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
174 	&VNET_NAME(icmptstamprepl), 0,
175 	"Respond to ICMP Timestamp packets");
176 
177 VNET_DEFINE_STATIC(int, error_keeptags) = 0;
178 #define	V_error_keeptags		VNET(error_keeptags)
179 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
180 	&VNET_NAME(error_keeptags), 0,
181 	"ICMP error response keeps copy of mbuf_tags of original packet");
182 
183 #ifdef ICMPPRINTFS
184 int	icmpprintfs = 0;
185 #endif
186 
187 static void	icmp_reflect(struct mbuf *);
188 static void	icmp_send(struct mbuf *, struct mbuf *);
189 static int	icmp_verify_redirect_gateway(struct sockaddr_in *,
190     struct sockaddr_in *, struct sockaddr_in *, u_int);
191 
192 extern	struct protosw inetsw[];
193 
194 /*
195  * Kernel module interface for updating icmpstat.  The argument is an index
196  * into icmpstat treated as an array of u_long.  While this encodes the
197  * general layout of icmpstat into the caller, it doesn't encode its
198  * location, so that future changes to add, for example, per-CPU stats
199  * support won't cause binary compatibility problems for kernel modules.
200  */
201 void
kmod_icmpstat_inc(int statnum)202 kmod_icmpstat_inc(int statnum)
203 {
204 
205 	counter_u64_add(VNET(icmpstat)[statnum], 1);
206 }
207 
208 /*
209  * Generate an error packet of type error
210  * in response to bad packet ip.
211  */
212 void
icmp_error(struct mbuf * n,int type,int code,uint32_t dest,int mtu)213 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
214 {
215 	struct ip *oip, *nip;
216 	struct icmp *icp;
217 	struct mbuf *m;
218 	unsigned icmplen, icmpelen, nlen, oiphlen;
219 
220 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
221 	    __func__));
222 
223 	if (type != ICMP_REDIRECT)
224 		ICMPSTAT_INC(icps_error);
225 	/*
226 	 * Don't send error:
227 	 *  if the original packet was encrypted.
228 	 *  if not the first fragment of message.
229 	 *  in response to a multicast or broadcast packet.
230 	 *  if the old packet protocol was an ICMP error message.
231 	 */
232 	if (n->m_flags & M_DECRYPTED)
233 		goto freeit;
234 	if (n->m_flags & (M_BCAST|M_MCAST))
235 		goto freeit;
236 
237 	/* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
238 	if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
239 		goto freeit;
240 	oip = mtod(n, struct ip *);
241 	oiphlen = oip->ip_hl << 2;
242 	if (n->m_len < oiphlen + ICMP_MINLEN)
243 		goto freeit;
244 #ifdef ICMPPRINTFS
245 	if (icmpprintfs)
246 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
247 #endif
248 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
249 		goto freeit;
250 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
251 	    !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
252 		oiphlen))->icmp_type)) {
253 		ICMPSTAT_INC(icps_oldicmp);
254 		goto freeit;
255 	}
256 	/*
257 	 * Calculate length to quote from original packet and
258 	 * prevent the ICMP mbuf from overflowing.
259 	 * Unfortunately this is non-trivial since ip_forward()
260 	 * sends us truncated packets.
261 	 */
262 	nlen = m_length(n, NULL);
263 	if (oip->ip_p == IPPROTO_TCP) {
264 		struct tcphdr *th;
265 		int tcphlen;
266 
267 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
268 		    n->m_next == NULL)
269 			goto stdreply;
270 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
271 		    (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
272 			goto freeit;
273 		oip = mtod(n, struct ip *);
274 		th = mtodo(n, oiphlen);
275 		tcphlen = th->th_off << 2;
276 		if (tcphlen < sizeof(struct tcphdr))
277 			goto freeit;
278 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
279 			goto freeit;
280 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
281 			goto stdreply;
282 		if (n->m_len < oiphlen + tcphlen &&
283 		    (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
284 			goto freeit;
285 		oip = mtod(n, struct ip *);
286 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
287 		    ntohs(oip->ip_len) - oiphlen));
288 	} else if (oip->ip_p == IPPROTO_SCTP) {
289 		struct sctphdr *sh;
290 		struct sctp_chunkhdr *ch;
291 
292 		if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
293 			goto stdreply;
294 		if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
295 		    n->m_next == NULL)
296 			goto stdreply;
297 		if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
298 		    (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
299 			goto freeit;
300 		oip = mtod(n, struct ip *);
301 		icmpelen = max(sizeof(struct sctphdr),
302 		    min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
303 		sh = mtodo(n, oiphlen);
304 		if (ntohl(sh->v_tag) == 0 &&
305 		    ntohs(oip->ip_len) >= oiphlen +
306 		    sizeof(struct sctphdr) + 8 &&
307 		    (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
308 		     n->m_next != NULL)) {
309 			if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
310 			    (n = m_pullup(n, oiphlen +
311 			    sizeof(struct sctphdr) + 8)) == NULL)
312 				goto freeit;
313 			oip = mtod(n, struct ip *);
314 			sh = mtodo(n, oiphlen);
315 			ch = (struct sctp_chunkhdr *)(sh + 1);
316 			if (ch->chunk_type == SCTP_INITIATION) {
317 				icmpelen = max(sizeof(struct sctphdr) + 8,
318 				    min(V_icmp_quotelen, ntohs(oip->ip_len) -
319 				    oiphlen));
320 			}
321 		}
322 	} else
323 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
324 		    oiphlen));
325 
326 	icmplen = min(oiphlen + icmpelen, nlen);
327 	if (icmplen < sizeof(struct ip))
328 		goto freeit;
329 
330 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
331 		m = m_gethdr(M_NOWAIT, MT_DATA);
332 	else
333 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
334 	if (m == NULL)
335 		goto freeit;
336 #ifdef MAC
337 	mac_netinet_icmp_reply(n, m);
338 #endif
339 	icmplen = min(icmplen, M_TRAILINGSPACE(m) -
340 	    sizeof(struct ip) - ICMP_MINLEN);
341 	m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
342 	m->m_data += sizeof(struct ip);
343 	m->m_len = ICMP_MINLEN + icmplen;
344 
345 	/* XXX MRT  make the outgoing packet use the same FIB
346 	 * that was associated with the incoming packet
347 	 */
348 	M_SETFIB(m, M_GETFIB(n));
349 	icp = mtod(m, struct icmp *);
350 	ICMPSTAT_INC(icps_outhist[type]);
351 	icp->icmp_type = type;
352 	if (type == ICMP_REDIRECT)
353 		icp->icmp_gwaddr.s_addr = dest;
354 	else {
355 		icp->icmp_void = 0;
356 		/*
357 		 * The following assignments assume an overlay with the
358 		 * just zeroed icmp_void field.
359 		 */
360 		if (type == ICMP_PARAMPROB) {
361 			icp->icmp_pptr = code;
362 			code = 0;
363 		} else if (type == ICMP_UNREACH &&
364 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
365 			icp->icmp_nextmtu = htons(mtu);
366 		}
367 	}
368 	icp->icmp_code = code;
369 
370 	/*
371 	 * Copy the quotation into ICMP message and
372 	 * convert quoted IP header back to network representation.
373 	 */
374 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
375 	nip = &icp->icmp_ip;
376 
377 	/*
378 	 * Set up ICMP message mbuf and copy old IP header (without options
379 	 * in front of ICMP message.
380 	 * If the original mbuf was meant to bypass the firewall, the error
381 	 * reply should bypass as well.
382 	 */
383 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
384 	KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
385 	    ("insufficient space for ip header"));
386 	m->m_data -= sizeof(struct ip);
387 	m->m_len += sizeof(struct ip);
388 	m->m_pkthdr.len = m->m_len;
389 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
390 	nip = mtod(m, struct ip *);
391 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
392 	nip->ip_len = htons(m->m_len);
393 	nip->ip_v = IPVERSION;
394 	nip->ip_hl = 5;
395 	nip->ip_p = IPPROTO_ICMP;
396 	nip->ip_tos = 0;
397 	nip->ip_off = 0;
398 
399 	if (V_error_keeptags)
400 		m_tag_copy_chain(m, n, M_NOWAIT);
401 
402 	icmp_reflect(m);
403 
404 freeit:
405 	m_freem(n);
406 }
407 
408 /*
409  * Process a received ICMP message.
410  */
411 int
icmp_input(struct mbuf ** mp,int * offp,int proto)412 icmp_input(struct mbuf **mp, int *offp, int proto)
413 {
414 	struct icmp *icp;
415 	struct in_ifaddr *ia;
416 	struct mbuf *m = *mp;
417 	struct ip *ip = mtod(m, struct ip *);
418 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
419 	int hlen = *offp;
420 	int icmplen = ntohs(ip->ip_len) - *offp;
421 	int i, code;
422 	void (*ctlfunc)(int, struct sockaddr *, void *);
423 	int fibnum;
424 
425 	NET_EPOCH_ASSERT();
426 
427 	*mp = NULL;
428 
429 	/*
430 	 * Locate icmp structure in mbuf, and check
431 	 * that not corrupted and of at least minimum length.
432 	 */
433 #ifdef ICMPPRINTFS
434 	if (icmpprintfs) {
435 		char srcbuf[INET_ADDRSTRLEN];
436 		char dstbuf[INET_ADDRSTRLEN];
437 
438 		printf("icmp_input from %s to %s, len %d\n",
439 		    inet_ntoa_r(ip->ip_src, srcbuf),
440 		    inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
441 	}
442 #endif
443 	if (icmplen < ICMP_MINLEN) {
444 		ICMPSTAT_INC(icps_tooshort);
445 		goto freeit;
446 	}
447 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
448 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
449 		ICMPSTAT_INC(icps_tooshort);
450 		return (IPPROTO_DONE);
451 	}
452 	ip = mtod(m, struct ip *);
453 	m->m_len -= hlen;
454 	m->m_data += hlen;
455 	icp = mtod(m, struct icmp *);
456 	if (in_cksum(m, icmplen)) {
457 		ICMPSTAT_INC(icps_checksum);
458 		goto freeit;
459 	}
460 	m->m_len += hlen;
461 	m->m_data -= hlen;
462 
463 #ifdef ICMPPRINTFS
464 	if (icmpprintfs)
465 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
466 		    icp->icmp_code);
467 #endif
468 
469 	/*
470 	 * Message type specific processing.
471 	 */
472 	if (icp->icmp_type > ICMP_MAXTYPE)
473 		goto raw;
474 
475 	/* Initialize */
476 	bzero(&icmpsrc, sizeof(icmpsrc));
477 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
478 	icmpsrc.sin_family = AF_INET;
479 	bzero(&icmpdst, sizeof(icmpdst));
480 	icmpdst.sin_len = sizeof(struct sockaddr_in);
481 	icmpdst.sin_family = AF_INET;
482 	bzero(&icmpgw, sizeof(icmpgw));
483 	icmpgw.sin_len = sizeof(struct sockaddr_in);
484 	icmpgw.sin_family = AF_INET;
485 
486 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
487 	code = icp->icmp_code;
488 	switch (icp->icmp_type) {
489 	case ICMP_UNREACH:
490 		switch (code) {
491 			case ICMP_UNREACH_NET:
492 			case ICMP_UNREACH_HOST:
493 			case ICMP_UNREACH_SRCFAIL:
494 			case ICMP_UNREACH_NET_UNKNOWN:
495 			case ICMP_UNREACH_HOST_UNKNOWN:
496 			case ICMP_UNREACH_ISOLATED:
497 			case ICMP_UNREACH_TOSNET:
498 			case ICMP_UNREACH_TOSHOST:
499 			case ICMP_UNREACH_HOST_PRECEDENCE:
500 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
501 				code = PRC_UNREACH_NET;
502 				break;
503 
504 			case ICMP_UNREACH_NEEDFRAG:
505 				code = PRC_MSGSIZE;
506 				break;
507 
508 			/*
509 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
510 			 * Treat subcodes 2,3 as immediate RST
511 			 */
512 			case ICMP_UNREACH_PROTOCOL:
513 				code = PRC_UNREACH_PROTOCOL;
514 				break;
515 			case ICMP_UNREACH_PORT:
516 				code = PRC_UNREACH_PORT;
517 				break;
518 
519 			case ICMP_UNREACH_NET_PROHIB:
520 			case ICMP_UNREACH_HOST_PROHIB:
521 			case ICMP_UNREACH_FILTER_PROHIB:
522 				code = PRC_UNREACH_ADMIN_PROHIB;
523 				break;
524 
525 			default:
526 				goto badcode;
527 		}
528 		goto deliver;
529 
530 	case ICMP_TIMXCEED:
531 		if (code > 1)
532 			goto badcode;
533 		code += PRC_TIMXCEED_INTRANS;
534 		goto deliver;
535 
536 	case ICMP_PARAMPROB:
537 		if (code > 1)
538 			goto badcode;
539 		code = PRC_PARAMPROB;
540 	deliver:
541 		/*
542 		 * Problem with datagram; advise higher level routines.
543 		 */
544 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
545 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
546 			ICMPSTAT_INC(icps_badlen);
547 			goto freeit;
548 		}
549 		/* Discard ICMP's in response to multicast packets */
550 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
551 			goto badcode;
552 #ifdef ICMPPRINTFS
553 		if (icmpprintfs)
554 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
555 #endif
556 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
557 		/*
558 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
559 		 * notification to TCP layer.
560 		 */
561 		i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
562 		ip_stripoptions(m);
563 		if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
564 			/* This should actually not happen */
565 			ICMPSTAT_INC(icps_tooshort);
566 			return (IPPROTO_DONE);
567 		}
568 		ip = mtod(m, struct ip *);
569 		icp = (struct icmp *)(ip + 1);
570 		/*
571 		 * The upper layer handler can rely on:
572 		 * - The outer IP header has no options.
573 		 * - The outer IP header, the ICMP header, the inner IP header,
574 		 *   and the first n bytes of the inner payload are contiguous.
575 		 *   n is at least 8, but might be larger based on
576 		 *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
577 		 */
578 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
579 		if (ctlfunc)
580 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
581 				   (void *)&icp->icmp_ip);
582 		break;
583 
584 	badcode:
585 		ICMPSTAT_INC(icps_badcode);
586 		break;
587 
588 	case ICMP_ECHO:
589 		if (!V_icmpbmcastecho
590 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
591 			ICMPSTAT_INC(icps_bmcastecho);
592 			break;
593 		}
594 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
595 			goto freeit;
596 		icp->icmp_type = ICMP_ECHOREPLY;
597 		goto reflect;
598 
599 	case ICMP_TSTAMP:
600 		if (V_icmptstamprepl == 0)
601 			break;
602 		if (!V_icmpbmcastecho
603 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
604 			ICMPSTAT_INC(icps_bmcasttstamp);
605 			break;
606 		}
607 		if (icmplen < ICMP_TSLEN) {
608 			ICMPSTAT_INC(icps_badlen);
609 			break;
610 		}
611 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
612 			goto freeit;
613 		icp->icmp_type = ICMP_TSTAMPREPLY;
614 		icp->icmp_rtime = iptime();
615 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
616 		goto reflect;
617 
618 	case ICMP_MASKREQ:
619 		if (V_icmpmaskrepl == 0)
620 			break;
621 		/*
622 		 * We are not able to respond with all ones broadcast
623 		 * unless we receive it over a point-to-point interface.
624 		 */
625 		if (icmplen < ICMP_MASKLEN)
626 			break;
627 		switch (ip->ip_dst.s_addr) {
628 		case INADDR_BROADCAST:
629 		case INADDR_ANY:
630 			icmpdst.sin_addr = ip->ip_src;
631 			break;
632 
633 		default:
634 			icmpdst.sin_addr = ip->ip_dst;
635 		}
636 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
637 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
638 		if (ia == NULL)
639 			break;
640 		if (ia->ia_ifp == NULL)
641 			break;
642 		icp->icmp_type = ICMP_MASKREPLY;
643 		if (V_icmpmaskfake == 0)
644 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
645 		else
646 			icp->icmp_mask = V_icmpmaskfake;
647 		if (ip->ip_src.s_addr == 0) {
648 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
649 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
650 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
651 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
652 		}
653 reflect:
654 		ICMPSTAT_INC(icps_reflect);
655 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
656 		icmp_reflect(m);
657 		return (IPPROTO_DONE);
658 
659 	case ICMP_REDIRECT:
660 		if (V_log_redirect) {
661 			u_long src, dst, gw;
662 
663 			src = ntohl(ip->ip_src.s_addr);
664 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
665 			gw = ntohl(icp->icmp_gwaddr.s_addr);
666 			printf("icmp redirect from %d.%d.%d.%d: "
667 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
668 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
669 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
670 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
671 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
672 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
673 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
674 		}
675 		/*
676 		 * RFC1812 says we must ignore ICMP redirects if we
677 		 * are acting as router.
678 		 */
679 		if (V_drop_redirect || V_ipforwarding)
680 			break;
681 		if (code > 3)
682 			goto badcode;
683 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
684 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
685 			ICMPSTAT_INC(icps_badlen);
686 			break;
687 		}
688 		/*
689 		 * Short circuit routing redirects to force
690 		 * immediate change in the kernel's routing
691 		 * tables.  The message is also handed to anyone
692 		 * listening on a raw socket (e.g. the routing
693 		 * daemon for use in updating its tables).
694 		 */
695 		icmpgw.sin_addr = ip->ip_src;
696 		icmpdst.sin_addr = icp->icmp_gwaddr;
697 #ifdef	ICMPPRINTFS
698 		if (icmpprintfs) {
699 			char dstbuf[INET_ADDRSTRLEN];
700 			char gwbuf[INET_ADDRSTRLEN];
701 
702 			printf("redirect dst %s to %s\n",
703 			       inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
704 			       inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
705 		}
706 #endif
707 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
708 
709 		/*
710 		 * RFC 1122 says network (code 0,2) redirects SHOULD
711 		 * be treated identically to the host redirects.
712 		 * Given that, ignore network masks.
713 		 */
714 
715 		/*
716 		 * Variable values:
717 		 * icmpsrc: route destination
718 		 * icmpdst: route gateway
719 		 * icmpgw: message source
720 		 */
721 
722 		if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst,
723 		    M_GETFIB(m)) != 0) {
724 			/* TODO: increment bad redirects here */
725 			break;
726 		}
727 
728 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
729 			rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc,
730 			    (struct sockaddr *)&icmpdst,
731 			    (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif,
732 			    RTF_GATEWAY, V_redirtimeout);
733 		}
734 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
735 		break;
736 
737 	/*
738 	 * No kernel processing for the following;
739 	 * just fall through to send to raw listener.
740 	 */
741 	case ICMP_ECHOREPLY:
742 	case ICMP_ROUTERADVERT:
743 	case ICMP_ROUTERSOLICIT:
744 	case ICMP_TSTAMPREPLY:
745 	case ICMP_IREQREPLY:
746 	case ICMP_MASKREPLY:
747 	case ICMP_SOURCEQUENCH:
748 	default:
749 		break;
750 	}
751 
752 raw:
753 	*mp = m;
754 	rip_input(mp, offp, proto);
755 	return (IPPROTO_DONE);
756 
757 freeit:
758 	m_freem(m);
759 	return (IPPROTO_DONE);
760 }
761 
762 /*
763  * Reflect the ip packet back to the source
764  */
765 static void
icmp_reflect(struct mbuf * m)766 icmp_reflect(struct mbuf *m)
767 {
768 	struct rm_priotracker in_ifa_tracker;
769 	struct ip *ip = mtod(m, struct ip *);
770 	struct ifaddr *ifa;
771 	struct ifnet *ifp;
772 	struct in_ifaddr *ia;
773 	struct in_addr t;
774 	struct nhop_object *nh;
775 	struct mbuf *opts = NULL;
776 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
777 
778 	NET_EPOCH_ASSERT();
779 
780 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
781 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
782 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
783 		m_freem(m);	/* Bad return address */
784 		ICMPSTAT_INC(icps_badaddr);
785 		goto done;	/* Ip_output() will check for broadcast */
786 	}
787 
788 	t = ip->ip_dst;
789 	ip->ip_dst = ip->ip_src;
790 
791 	/*
792 	 * Source selection for ICMP replies:
793 	 *
794 	 * If the incoming packet was addressed directly to one of our
795 	 * own addresses, use dst as the src for the reply.
796 	 */
797 	IN_IFADDR_RLOCK(&in_ifa_tracker);
798 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
799 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
800 			t = IA_SIN(ia)->sin_addr;
801 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
802 			goto match;
803 		}
804 	}
805 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
806 
807 	/*
808 	 * If the incoming packet was addressed to one of our broadcast
809 	 * addresses, use the first non-broadcast address which corresponds
810 	 * to the incoming interface.
811 	 */
812 	ifp = m->m_pkthdr.rcvif;
813 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
814 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
815 			if (ifa->ifa_addr->sa_family != AF_INET)
816 				continue;
817 			ia = ifatoia(ifa);
818 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
819 			    t.s_addr) {
820 				t = IA_SIN(ia)->sin_addr;
821 				goto match;
822 			}
823 		}
824 	}
825 	/*
826 	 * If the packet was transiting through us, use the address of
827 	 * the interface the packet came through in.  If that interface
828 	 * doesn't have a suitable IP address, the normal selection
829 	 * criteria apply.
830 	 */
831 	if (V_icmp_rfi && ifp != NULL) {
832 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
833 			if (ifa->ifa_addr->sa_family != AF_INET)
834 				continue;
835 			ia = ifatoia(ifa);
836 			t = IA_SIN(ia)->sin_addr;
837 			goto match;
838 		}
839 	}
840 	/*
841 	 * If the incoming packet was not addressed directly to us, use
842 	 * designated interface for icmp replies specified by sysctl
843 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
844 	 * with normal source selection.
845 	 */
846 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
847 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
848 			if (ifa->ifa_addr->sa_family != AF_INET)
849 				continue;
850 			ia = ifatoia(ifa);
851 			t = IA_SIN(ia)->sin_addr;
852 			goto match;
853 		}
854 	}
855 	/*
856 	 * If the packet was transiting through us, use the address of
857 	 * the interface that is the closest to the packet source.
858 	 * When we don't have a route back to the packet source, stop here
859 	 * and drop the packet.
860 	 */
861 	nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
862 	if (nh == NULL) {
863 		m_freem(m);
864 		ICMPSTAT_INC(icps_noroute);
865 		goto done;
866 	}
867 	t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr;
868 match:
869 #ifdef MAC
870 	mac_netinet_icmp_replyinplace(m);
871 #endif
872 	ip->ip_src = t;
873 	ip->ip_ttl = V_ip_defttl;
874 
875 	if (optlen > 0) {
876 		u_char *cp;
877 		int opt, cnt;
878 		u_int len;
879 
880 		/*
881 		 * Retrieve any source routing from the incoming packet;
882 		 * add on any record-route or timestamp options.
883 		 */
884 		cp = (u_char *) (ip + 1);
885 		if ((opts = ip_srcroute(m)) == NULL &&
886 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
887 			opts->m_len = sizeof(struct in_addr);
888 			mtod(opts, struct in_addr *)->s_addr = 0;
889 		}
890 		if (opts) {
891 #ifdef ICMPPRINTFS
892 		    if (icmpprintfs)
893 			    printf("icmp_reflect optlen %d rt %d => ",
894 				optlen, opts->m_len);
895 #endif
896 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
897 			    opt = cp[IPOPT_OPTVAL];
898 			    if (opt == IPOPT_EOL)
899 				    break;
900 			    if (opt == IPOPT_NOP)
901 				    len = 1;
902 			    else {
903 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
904 					    break;
905 				    len = cp[IPOPT_OLEN];
906 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
907 				        len > cnt)
908 					    break;
909 			    }
910 			    /*
911 			     * Should check for overflow, but it "can't happen"
912 			     */
913 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
914 				opt == IPOPT_SECURITY) {
915 				    bcopy((caddr_t)cp,
916 					mtod(opts, caddr_t) + opts->m_len, len);
917 				    opts->m_len += len;
918 			    }
919 		    }
920 		    /* Terminate & pad, if necessary */
921 		    cnt = opts->m_len % 4;
922 		    if (cnt) {
923 			    for (; cnt < 4; cnt++) {
924 				    *(mtod(opts, caddr_t) + opts->m_len) =
925 					IPOPT_EOL;
926 				    opts->m_len++;
927 			    }
928 		    }
929 #ifdef ICMPPRINTFS
930 		    if (icmpprintfs)
931 			    printf("%d\n", opts->m_len);
932 #endif
933 		}
934 		ip_stripoptions(m);
935 	}
936 	m_tag_delete_nonpersistent(m);
937 	m->m_flags &= ~(M_BCAST|M_MCAST);
938 	icmp_send(m, opts);
939 done:
940 	if (opts)
941 		(void)m_free(opts);
942 }
943 
944 /*
945  * Verifies if redirect message is valid, according to RFC 1122
946  *
947  * @src: sockaddr with address of redirect originator
948  * @dst: sockaddr with destination in question
949  * @gateway: new proposed gateway
950  *
951  * Returns 0 on success.
952  */
953 static int
icmp_verify_redirect_gateway(struct sockaddr_in * src,struct sockaddr_in * dst,struct sockaddr_in * gateway,u_int fibnum)954 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst,
955     struct sockaddr_in *gateway, u_int fibnum)
956 {
957 	struct nhop_object *nh;
958 	struct ifaddr *ifa;
959 
960 	NET_EPOCH_ASSERT();
961 
962 	/* Verify the gateway is directly reachable. */
963 	if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL)
964 		return (ENETUNREACH);
965 
966 	/* TODO: fib-aware. */
967 	if (ifa_ifwithaddr_check((struct sockaddr *)gateway))
968 		return (EHOSTUNREACH);
969 
970 	nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0);
971 	if (nh == NULL)
972 		return (EINVAL);
973 
974 	/*
975 	 * If the redirect isn't from our current router for this dst,
976 	 * it's either old or wrong.  If it redirects us to ourselves,
977 	 * we have a routing loop, perhaps as a result of an interface
978 	 * going down recently.
979 	 */
980 	if (!sa_equal((struct sockaddr *)src, &nh->gw_sa))
981 		return (EINVAL);
982 	if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK)
983 		return (EINVAL);
984 
985 	/* If host route already exists, ignore redirect. */
986 	if (nh->nh_flags & NHF_HOST)
987 		return (EEXIST);
988 
989 	/* If the prefix is directly reachable, ignore redirect. */
990 	if (!(nh->nh_flags & NHF_GATEWAY))
991 		return (EEXIST);
992 
993 	return (0);
994 }
995 
996 /*
997  * Send an icmp packet back to the ip level,
998  * after supplying a checksum.
999  */
1000 static void
icmp_send(struct mbuf * m,struct mbuf * opts)1001 icmp_send(struct mbuf *m, struct mbuf *opts)
1002 {
1003 	struct ip *ip = mtod(m, struct ip *);
1004 	int hlen;
1005 	struct icmp *icp;
1006 
1007 	hlen = ip->ip_hl << 2;
1008 	m->m_data += hlen;
1009 	m->m_len -= hlen;
1010 	icp = mtod(m, struct icmp *);
1011 	icp->icmp_cksum = 0;
1012 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
1013 	m->m_data -= hlen;
1014 	m->m_len += hlen;
1015 	m->m_pkthdr.rcvif = (struct ifnet *)0;
1016 #ifdef ICMPPRINTFS
1017 	if (icmpprintfs) {
1018 		char dstbuf[INET_ADDRSTRLEN];
1019 		char srcbuf[INET_ADDRSTRLEN];
1020 
1021 		printf("icmp_send dst %s src %s\n",
1022 		    inet_ntoa_r(ip->ip_dst, dstbuf),
1023 		    inet_ntoa_r(ip->ip_src, srcbuf));
1024 	}
1025 #endif
1026 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
1027 }
1028 
1029 /*
1030  * Return milliseconds since 00:00 UTC in network format.
1031  */
1032 uint32_t
iptime(void)1033 iptime(void)
1034 {
1035 	struct timeval atv;
1036 	u_long t;
1037 
1038 	getmicrotime(&atv);
1039 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1040 	return (htonl(t));
1041 }
1042 
1043 /*
1044  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1045  * given current value MTU.  If DIR is less than zero, a larger plateau
1046  * is returned; otherwise, a smaller value is returned.
1047  */
1048 int
ip_next_mtu(int mtu,int dir)1049 ip_next_mtu(int mtu, int dir)
1050 {
1051 	static int mtutab[] = {
1052 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
1053 		296, 68, 0
1054 	};
1055 	int i, size;
1056 
1057 	size = (sizeof mtutab) / (sizeof mtutab[0]);
1058 	if (dir >= 0) {
1059 		for (i = 0; i < size; i++)
1060 			if (mtu > mtutab[i])
1061 				return mtutab[i];
1062 	} else {
1063 		for (i = size - 1; i >= 0; i--)
1064 			if (mtu < mtutab[i])
1065 				return mtutab[i];
1066 		if (mtu == mtutab[0])
1067 			return mtutab[0];
1068 	}
1069 	return 0;
1070 }
1071 #endif /* INET */
1072 
1073 /*
1074  * badport_bandlim() - check for ICMP bandwidth limit
1075  *
1076  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
1077  *	hit our bandwidth limit and it is not ok.
1078  *
1079  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
1080  *
1081  *	For now we separate the TCP and UDP subsystems w/ different 'which'
1082  *	values.  We may eventually remove this separation (and simplify the
1083  *	code further).
1084  *
1085  *	Note that the printing of the error message is delayed so we can
1086  *	properly print the icmp error rate that the system was trying to do
1087  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
1088  *	the 'final' error, but it doesn't make sense to solve the printing
1089  *	delay with more complex code.
1090  */
1091 VNET_DEFINE_STATIC(struct counter_rate, icmp_rates[BANDLIM_MAX]);
1092 #define	V_icmp_rates	VNET(icmp_rates)
1093 
1094 static const char *icmp_rate_descrs[BANDLIM_MAX] = {
1095 	[BANDLIM_ICMP_UNREACH] = "icmp unreach",
1096 	[BANDLIM_ICMP_ECHO] = "icmp ping",
1097 	[BANDLIM_ICMP_TSTAMP] = "icmp tstamp",
1098 	[BANDLIM_RST_CLOSEDPORT] = "closed port RST",
1099 	[BANDLIM_RST_OPENPORT] = "open port RST",
1100 	[BANDLIM_ICMP6_UNREACH] = "icmp6 unreach",
1101 	[BANDLIM_SCTP_OOTB] = "sctp ootb",
1102 };
1103 
1104 static void
icmplim_new_jitter(int which)1105 icmplim_new_jitter(int which)
1106 {
1107 	/*
1108 	 * Adjust limit +/- to jitter the measurement to deny a side-channel
1109 	 * port scan as in https://dl.acm.org/doi/10.1145/3372297.3417280
1110 	 */
1111 	KASSERT(which >= 0 && which < BANDLIM_MAX,
1112 	    ("%s: which %d", __func__, which));
1113 	if (V_icmplim_jitter > 0)
1114 		V_icmplim_curr_jitter[which] =
1115 		    arc4random_uniform(V_icmplim_jitter * 2 + 1) -
1116 		    V_icmplim_jitter;
1117 }
1118 
1119 static int
sysctl_icmplim_and_jitter(SYSCTL_HANDLER_ARGS)1120 sysctl_icmplim_and_jitter(SYSCTL_HANDLER_ARGS)
1121 {
1122 	uint32_t new;
1123 	int error;
1124 	bool lim;
1125 
1126 	MPASS(oidp->oid_arg1 == &VNET_NAME(icmplim) ||
1127 	    oidp->oid_arg1 == &VNET_NAME(icmplim_jitter));
1128 
1129 	lim = (oidp->oid_arg1 == &VNET_NAME(icmplim));
1130 	new = lim ? V_icmplim : V_icmplim_jitter;
1131 	error = sysctl_handle_int(oidp, &new, 0, req);
1132 	if (error == 0 && req->newptr) {
1133 		if (lim) {
1134 			if (new != 0 && new <= V_icmplim_jitter)
1135 				error = EINVAL;
1136 			else
1137 				V_icmplim = new;
1138 		} else {
1139 			if (new >= V_icmplim)
1140 				error = EINVAL;
1141 			else {
1142 				V_icmplim_jitter = new;
1143 				for (int i = 0; i < BANDLIM_MAX; i++) {
1144 					icmplim_new_jitter(i);
1145 				}
1146 			}
1147 		}
1148 	}
1149 	MPASS(V_icmplim == 0 || V_icmplim > V_icmplim_jitter);
1150 
1151 	return (error);
1152 }
1153 
1154 static void
icmp_bandlimit_init(void)1155 icmp_bandlimit_init(void)
1156 {
1157 
1158 	for (int i = 0; i < BANDLIM_MAX; i++) {
1159 		V_icmp_rates[i].cr_rate = counter_u64_alloc(M_WAITOK);
1160 		V_icmp_rates[i].cr_ticks = ticks;
1161 		icmplim_new_jitter(i);
1162 	}
1163 }
1164 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1165     icmp_bandlimit_init, NULL);
1166 
1167 #ifdef VIMAGE
1168 static void
icmp_bandlimit_uninit(void)1169 icmp_bandlimit_uninit(void)
1170 {
1171 
1172 	for (int i = 0; i < BANDLIM_MAX; i++)
1173 		counter_u64_free(V_icmp_rates[i].cr_rate);
1174 }
1175 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1176     icmp_bandlimit_uninit, NULL);
1177 #endif
1178 
1179 int
badport_bandlim(int which)1180 badport_bandlim(int which)
1181 {
1182 	int64_t pps;
1183 
1184 	if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1185 		return (0);
1186 
1187 	KASSERT(which >= 0 && which < BANDLIM_MAX,
1188 	    ("%s: which %d", __func__, which));
1189 
1190 	pps = counter_ratecheck(&V_icmp_rates[which], V_icmplim +
1191 	    V_icmplim_curr_jitter[which]);
1192 	if (pps > 0) {
1193 		if (V_icmplim_output)
1194 			log(LOG_NOTICE,
1195 			    "Limiting %s response from %jd to %d packets/sec\n",
1196 			    icmp_rate_descrs[which], (intmax_t )pps,
1197 			    V_icmplim + V_icmplim_curr_jitter[which]);
1198 		icmplim_new_jitter(which);
1199 	}
1200 	if (pps == -1)
1201 		return (-1);
1202 	return (0);
1203 }
1204