2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
31 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
39 #include <sys/param.h>
40 #include <sys/systm.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
45 #include <sys/kernel.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
51 #include <net/if_var.h>
52 #include <net/if_private.h>
53 #include <net/if_types.h>
54 #include <net/route.h>
55 #include <net/route/route_ctl.h>
56 #include <net/route/nhop.h>
59 #include <netinet/in.h>
60 #include <netinet/in_fib.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #include <netinet/ip_icmp.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/ip_options.h>
68 #include <netinet/sctp.h>
69 #include <netinet/tcp.h>
70 #include <netinet/tcpip.h>
71 #include <netinet/icmp_var.h>
75 #include <machine/in_cksum.h>
77 #include <security/mac/mac_framework.h>
80 extern ipproto_ctlinput_t *ip_ctlprotox[];
83 * ICMP routines: error generation, receive packet processing, and
84 * routines to turnaround packets back to the originator, and
85 * host table maintenance routines.
87 VNET_DEFINE_STATIC(int, icmplim) = 200;
88 #define V_icmplim VNET(icmplim)
89 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
90 &VNET_NAME(icmplim), 0,
91 "Maximum number of ICMP responses per second");
93 VNET_DEFINE_STATIC(int, icmplim_curr_jitter) = 0;
94 #define V_icmplim_curr_jitter VNET(icmplim_curr_jitter)
95 VNET_DEFINE_STATIC(int, icmplim_jitter) = 16;
96 #define V_icmplim_jitter VNET(icmplim_jitter)
97 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLFLAG_VNET | CTLFLAG_RW,
98 &VNET_NAME(icmplim_jitter), 0,
99 "Random icmplim jitter adjustment limit");
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");
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)");
114 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
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");
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");
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");
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");
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");
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");
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");
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");
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");
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");
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);
193 * Kernel module interface for updating icmpstat. The argument is an index
194 * into icmpstat treated as an array of u_long. While this encodes the
195 * general layout of icmpstat into the caller, it doesn't encode its
196 * location, so that future changes to add, for example, per-CPU stats
197 * support won't cause binary compatibility problems for kernel modules.
200 kmod_icmpstat_inc(int statnum)
203 counter_u64_add(VNET(icmpstat)[statnum], 1);
207 * Generate an error packet of type error
208 * in response to bad packet ip.
211 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
213 struct ip *oip, *nip;
216 unsigned icmplen, icmpelen, nlen, oiphlen;
218 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
221 if (type != ICMP_REDIRECT)
222 ICMPSTAT_INC(icps_error);
225 * if the original packet was encrypted.
226 * if not the first fragment of message.
227 * in response to a multicast or broadcast packet.
228 * if the old packet protocol was an ICMP error message.
230 if (n->m_flags & M_DECRYPTED)
232 if (n->m_flags & (M_BCAST|M_MCAST))
235 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
236 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
238 oip = mtod(n, struct ip *);
239 oiphlen = oip->ip_hl << 2;
240 if (n->m_len < oiphlen + ICMP_MINLEN)
244 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
246 if (oip->ip_off & htons(~(IP_MF|IP_DF)))
248 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
249 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
250 oiphlen))->icmp_type)) {
251 ICMPSTAT_INC(icps_oldicmp);
255 * Calculate length to quote from original packet and
256 * prevent the ICMP mbuf from overflowing.
257 * Unfortunately this is non-trivial since ip_forward()
258 * sends us truncated packets.
260 nlen = m_length(n, NULL);
261 if (oip->ip_p == IPPROTO_TCP) {
265 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
268 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
269 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
271 oip = mtod(n, struct ip *);
272 th = mtodo(n, oiphlen);
273 tcphlen = th->th_off << 2;
274 if (tcphlen < sizeof(struct tcphdr))
276 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
278 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
280 if (n->m_len < oiphlen + tcphlen &&
281 (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
283 oip = mtod(n, struct ip *);
284 icmpelen = max(tcphlen, min(V_icmp_quotelen,
285 ntohs(oip->ip_len) - oiphlen));
286 } else if (oip->ip_p == IPPROTO_SCTP) {
288 struct sctp_chunkhdr *ch;
290 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
292 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
295 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
296 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
298 oip = mtod(n, struct ip *);
299 icmpelen = max(sizeof(struct sctphdr),
300 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
301 sh = mtodo(n, oiphlen);
302 if (ntohl(sh->v_tag) == 0 &&
303 ntohs(oip->ip_len) >= oiphlen +
304 sizeof(struct sctphdr) + 8 &&
305 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
306 n->m_next != NULL)) {
307 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
308 (n = m_pullup(n, oiphlen +
309 sizeof(struct sctphdr) + 8)) == NULL)
311 oip = mtod(n, struct ip *);
312 sh = mtodo(n, oiphlen);
313 ch = (struct sctp_chunkhdr *)(sh + 1);
314 if (ch->chunk_type == SCTP_INITIATION) {
315 icmpelen = max(sizeof(struct sctphdr) + 8,
316 min(V_icmp_quotelen, ntohs(oip->ip_len) -
321 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
324 icmplen = min(oiphlen + icmpelen, nlen);
325 if (icmplen < sizeof(struct ip))
328 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
329 m = m_gethdr(M_NOWAIT, MT_DATA);
331 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
335 mac_netinet_icmp_reply(n, m);
337 icmplen = min(icmplen, M_TRAILINGSPACE(m) -
338 sizeof(struct ip) - ICMP_MINLEN);
339 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
340 m->m_data += sizeof(struct ip);
341 m->m_len = ICMP_MINLEN + icmplen;
343 /* XXX MRT make the outgoing packet use the same FIB
344 * that was associated with the incoming packet
346 M_SETFIB(m, M_GETFIB(n));
347 icp = mtod(m, struct icmp *);
348 ICMPSTAT_INC(icps_outhist[type]);
349 icp->icmp_type = type;
350 if (type == ICMP_REDIRECT)
351 icp->icmp_gwaddr.s_addr = dest;
355 * The following assignments assume an overlay with the
356 * just zeroed icmp_void field.
358 if (type == ICMP_PARAMPROB) {
359 icp->icmp_pptr = code;
361 } else if (type == ICMP_UNREACH &&
362 code == ICMP_UNREACH_NEEDFRAG && mtu) {
363 icp->icmp_nextmtu = htons(mtu);
366 icp->icmp_code = code;
369 * Copy the quotation into ICMP message and
370 * convert quoted IP header back to network representation.
372 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
376 * Set up ICMP message mbuf and copy old IP header (without options
377 * in front of ICMP message.
378 * If the original mbuf was meant to bypass the firewall, the error
379 * reply should bypass as well.
381 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
382 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
383 ("insufficient space for ip header"));
384 m->m_data -= sizeof(struct ip);
385 m->m_len += sizeof(struct ip);
386 m->m_pkthdr.len = m->m_len;
387 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
388 nip = mtod(m, struct ip *);
389 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
390 nip->ip_len = htons(m->m_len);
391 nip->ip_v = IPVERSION;
393 nip->ip_p = IPPROTO_ICMP;
397 if (V_error_keeptags)
398 m_tag_copy_chain(m, n, M_NOWAIT);
407 icmp_errmap(const struct icmp *icp)
410 switch (icp->icmp_type) {
412 switch (icp->icmp_code) {
413 case ICMP_UNREACH_NET:
414 case ICMP_UNREACH_HOST:
415 case ICMP_UNREACH_SRCFAIL:
416 case ICMP_UNREACH_NET_UNKNOWN:
417 case ICMP_UNREACH_HOST_UNKNOWN:
418 case ICMP_UNREACH_ISOLATED:
419 case ICMP_UNREACH_TOSNET:
420 case ICMP_UNREACH_TOSHOST:
421 case ICMP_UNREACH_HOST_PRECEDENCE:
422 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
423 return (EHOSTUNREACH);
424 case ICMP_UNREACH_NEEDFRAG:
426 case ICMP_UNREACH_PROTOCOL:
427 case ICMP_UNREACH_PORT:
428 case ICMP_UNREACH_NET_PROHIB:
429 case ICMP_UNREACH_HOST_PROHIB:
430 case ICMP_UNREACH_FILTER_PROHIB:
431 return (ECONNREFUSED);
436 switch (icp->icmp_code) {
437 case ICMP_TIMXCEED_INTRANS:
438 return (EHOSTUNREACH);
443 switch (icp->icmp_code) {
444 case ICMP_PARAMPROB_ERRATPTR:
445 case ICMP_PARAMPROB_OPTABSENT:
446 return (ENOPROTOOPT);
456 * Process a received ICMP message.
459 icmp_input(struct mbuf **mp, int *offp, int proto)
462 struct in_ifaddr *ia;
463 struct mbuf *m = *mp;
464 struct ip *ip = mtod(m, struct ip *);
465 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
467 int icmplen = ntohs(ip->ip_len) - *offp;
476 * Locate icmp structure in mbuf, and check
477 * that not corrupted and of at least minimum length.
481 char srcbuf[INET_ADDRSTRLEN];
482 char dstbuf[INET_ADDRSTRLEN];
484 printf("icmp_input from %s to %s, len %d\n",
485 inet_ntoa_r(ip->ip_src, srcbuf),
486 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
489 if (icmplen < ICMP_MINLEN) {
490 ICMPSTAT_INC(icps_tooshort);
493 i = hlen + min(icmplen, ICMP_ADVLENMIN);
494 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
495 ICMPSTAT_INC(icps_tooshort);
496 return (IPPROTO_DONE);
498 ip = mtod(m, struct ip *);
501 icp = mtod(m, struct icmp *);
502 if (in_cksum(m, icmplen)) {
503 ICMPSTAT_INC(icps_checksum);
511 printf("icmp_input, type %d code %d\n", icp->icmp_type,
516 * Message type specific processing.
518 if (icp->icmp_type > ICMP_MAXTYPE)
522 bzero(&icmpsrc, sizeof(icmpsrc));
523 icmpsrc.sin_len = sizeof(struct sockaddr_in);
524 icmpsrc.sin_family = AF_INET;
525 bzero(&icmpdst, sizeof(icmpdst));
526 icmpdst.sin_len = sizeof(struct sockaddr_in);
527 icmpdst.sin_family = AF_INET;
528 bzero(&icmpgw, sizeof(icmpgw));
529 icmpgw.sin_len = sizeof(struct sockaddr_in);
530 icmpgw.sin_family = AF_INET;
532 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
533 code = icp->icmp_code;
534 switch (icp->icmp_type) {
536 if (code > ICMP_UNREACH_PRECEDENCE_CUTOFF)
542 if (code > ICMP_TIMXCEED_REASS)
548 if (code > ICMP_PARAMPROB_LENGTH)
553 * Problem with datagram; advise higher level routines.
555 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
556 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
557 ICMPSTAT_INC(icps_badlen);
560 /* Discard ICMP's in response to multicast packets */
561 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
563 /* Filter out responses to INADDR_ANY, protocols ignore it. */
564 if (icp->icmp_ip.ip_dst.s_addr == INADDR_ANY ||
565 icp->icmp_ip.ip_src.s_addr == INADDR_ANY)
569 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
572 * XXX if the packet contains [IPv4 AH TCP], we can't make a
573 * notification to TCP layer.
575 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
577 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
578 /* This should actually not happen */
579 ICMPSTAT_INC(icps_tooshort);
580 return (IPPROTO_DONE);
582 ip = mtod(m, struct ip *);
583 icp = (struct icmp *)(ip + 1);
585 * The upper layer handler can rely on:
586 * - The outer IP header has no options.
587 * - The outer IP header, the ICMP header, the inner IP header,
588 * and the first n bytes of the inner payload are contiguous.
589 * n is at least 8, but might be larger based on
590 * ICMP_ADVLENPREF. See its definition in ip_icmp.h.
592 if (ip_ctlprotox[icp->icmp_ip.ip_p] != NULL)
593 ip_ctlprotox[icp->icmp_ip.ip_p](icp);
597 ICMPSTAT_INC(icps_badcode);
601 if (!V_icmpbmcastecho
602 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
603 ICMPSTAT_INC(icps_bmcastecho);
606 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
608 icp->icmp_type = ICMP_ECHOREPLY;
612 if (V_icmptstamprepl == 0)
614 if (!V_icmpbmcastecho
615 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
616 ICMPSTAT_INC(icps_bmcasttstamp);
619 if (icmplen < ICMP_TSLEN) {
620 ICMPSTAT_INC(icps_badlen);
623 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
625 icp->icmp_type = ICMP_TSTAMPREPLY;
626 icp->icmp_rtime = iptime();
627 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
631 if (V_icmpmaskrepl == 0)
634 * We are not able to respond with all ones broadcast
635 * unless we receive it over a point-to-point interface.
637 if (icmplen < ICMP_MASKLEN)
639 switch (ip->ip_dst.s_addr) {
640 case INADDR_BROADCAST:
642 icmpdst.sin_addr = ip->ip_src;
646 icmpdst.sin_addr = ip->ip_dst;
648 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
649 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
652 if (ia->ia_ifp == NULL)
654 icp->icmp_type = ICMP_MASKREPLY;
655 if (V_icmpmaskfake == 0)
656 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
658 icp->icmp_mask = V_icmpmaskfake;
659 if (ip->ip_src.s_addr == 0) {
660 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
661 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
662 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
663 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
666 ICMPSTAT_INC(icps_reflect);
667 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
669 return (IPPROTO_DONE);
672 if (V_log_redirect) {
675 src = ntohl(ip->ip_src.s_addr);
676 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
677 gw = ntohl(icp->icmp_gwaddr.s_addr);
678 printf("icmp redirect from %d.%d.%d.%d: "
679 "%d.%d.%d.%d => %d.%d.%d.%d\n",
680 (int)(src >> 24), (int)((src >> 16) & 0xff),
681 (int)((src >> 8) & 0xff), (int)(src & 0xff),
682 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
683 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
684 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
685 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
688 * RFC1812 says we must ignore ICMP redirects if we
689 * are acting as router.
691 if (V_drop_redirect || V_ipforwarding)
695 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
696 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
697 ICMPSTAT_INC(icps_badlen);
701 * Short circuit routing redirects to force
702 * immediate change in the kernel's routing
703 * tables. The message is also handed to anyone
704 * listening on a raw socket (e.g. the routing
705 * daemon for use in updating its tables).
707 icmpgw.sin_addr = ip->ip_src;
708 icmpdst.sin_addr = icp->icmp_gwaddr;
711 char dstbuf[INET_ADDRSTRLEN];
712 char gwbuf[INET_ADDRSTRLEN];
714 printf("redirect dst %s to %s\n",
715 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
716 inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
719 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
722 * RFC 1122 says network (code 0,2) redirects SHOULD
723 * be treated identically to the host redirects.
724 * Given that, ignore network masks.
729 * icmpsrc: route destination
730 * icmpdst: route gateway
731 * icmpgw: message source
734 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst,
736 /* TODO: increment bad redirects here */
740 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
741 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc,
742 (struct sockaddr *)&icmpdst,
743 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif,
744 RTF_GATEWAY, V_redirtimeout);
749 * No kernel processing for the following;
750 * just fall through to send to raw listener.
753 case ICMP_ROUTERADVERT:
754 case ICMP_ROUTERSOLICIT:
755 case ICMP_TSTAMPREPLY:
758 case ICMP_SOURCEQUENCH:
765 rip_input(mp, offp, proto);
766 return (IPPROTO_DONE);
770 return (IPPROTO_DONE);
774 * Reflect the ip packet back to the source
777 icmp_reflect(struct mbuf *m)
779 struct ip *ip = mtod(m, struct ip *);
782 struct in_ifaddr *ia;
784 struct nhop_object *nh;
785 struct mbuf *opts = NULL;
786 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
790 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
791 (IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net240) ||
792 (IN_ZERONET(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net0) ) {
793 m_freem(m); /* Bad return address */
794 ICMPSTAT_INC(icps_badaddr);
795 goto done; /* Ip_output() will check for broadcast */
799 ip->ip_dst = ip->ip_src;
802 * Source selection for ICMP replies:
804 * If the incoming packet was addressed directly to one of our
805 * own addresses, use dst as the src for the reply.
807 CK_LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
808 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
809 t = IA_SIN(ia)->sin_addr;
815 * If the incoming packet was addressed to one of our broadcast
816 * addresses, use the first non-broadcast address which corresponds
817 * to the incoming interface.
819 ifp = m->m_pkthdr.rcvif;
820 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
821 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
822 if (ifa->ifa_addr->sa_family != AF_INET)
825 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
827 t = IA_SIN(ia)->sin_addr;
833 * If the packet was transiting through us, use the address of
834 * the interface the packet came through in. If that interface
835 * doesn't have a suitable IP address, the normal selection
838 if (V_icmp_rfi && ifp != NULL) {
839 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
840 if (ifa->ifa_addr->sa_family != AF_INET)
843 t = IA_SIN(ia)->sin_addr;
848 * If the incoming packet was not addressed directly to us, use
849 * designated interface for icmp replies specified by sysctl
850 * net.inet.icmp.reply_src (default not set). Otherwise continue
851 * with normal source selection.
853 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
854 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
855 if (ifa->ifa_addr->sa_family != AF_INET)
858 t = IA_SIN(ia)->sin_addr;
863 * If the packet was transiting through us, use the address of
864 * the interface that is the closest to the packet source.
865 * When we don't have a route back to the packet source, stop here
866 * and drop the packet.
868 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
871 ICMPSTAT_INC(icps_noroute);
874 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr;
877 mac_netinet_icmp_replyinplace(m);
880 ip->ip_ttl = V_ip_defttl;
888 * Retrieve any source routing from the incoming packet;
889 * add on any record-route or timestamp options.
891 cp = (u_char *) (ip + 1);
892 if ((opts = ip_srcroute(m)) == NULL &&
893 (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
894 opts->m_len = sizeof(struct in_addr);
895 mtod(opts, struct in_addr *)->s_addr = 0;
900 printf("icmp_reflect optlen %d rt %d => ",
901 optlen, opts->m_len);
903 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
904 opt = cp[IPOPT_OPTVAL];
905 if (opt == IPOPT_EOL)
907 if (opt == IPOPT_NOP)
910 if (cnt < IPOPT_OLEN + sizeof(*cp))
912 len = cp[IPOPT_OLEN];
913 if (len < IPOPT_OLEN + sizeof(*cp) ||
918 * Should check for overflow, but it "can't happen"
920 if (opt == IPOPT_RR || opt == IPOPT_TS ||
921 opt == IPOPT_SECURITY) {
923 mtod(opts, caddr_t) + opts->m_len, len);
927 /* Terminate & pad, if necessary */
928 cnt = opts->m_len % 4;
930 for (; cnt < 4; cnt++) {
931 *(mtod(opts, caddr_t) + opts->m_len) =
938 printf("%d\n", opts->m_len);
943 m_tag_delete_nonpersistent(m);
944 m->m_flags &= ~(M_BCAST|M_MCAST);
952 * Verifies if redirect message is valid, according to RFC 1122
954 * @src: sockaddr with address of redirect originator
955 * @dst: sockaddr with destination in question
956 * @gateway: new proposed gateway
958 * Returns 0 on success.
961 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst,
962 struct sockaddr_in *gateway, u_int fibnum)
964 struct nhop_object *nh;
969 /* Verify the gateway is directly reachable. */
970 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL)
971 return (ENETUNREACH);
973 /* TODO: fib-aware. */
974 if (ifa_ifwithaddr_check((struct sockaddr *)gateway))
975 return (EHOSTUNREACH);
977 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0);
982 * If the redirect isn't from our current router for this dst,
983 * it's either old or wrong. If it redirects us to ourselves,
984 * we have a routing loop, perhaps as a result of an interface
985 * going down recently.
987 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa))
989 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK)
992 /* If host route already exists, ignore redirect. */
993 if (nh->nh_flags & NHF_HOST)
996 /* If the prefix is directly reachable, ignore redirect. */
997 if (!(nh->nh_flags & NHF_GATEWAY))
1004 * Send an icmp packet back to the ip level,
1005 * after supplying a checksum.
1008 icmp_send(struct mbuf *m, struct mbuf *opts)
1010 struct ip *ip = mtod(m, struct ip *);
1014 hlen = ip->ip_hl << 2;
1017 icp = mtod(m, struct icmp *);
1018 icp->icmp_cksum = 0;
1019 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
1022 m->m_pkthdr.rcvif = (struct ifnet *)0;
1025 char dstbuf[INET_ADDRSTRLEN];
1026 char srcbuf[INET_ADDRSTRLEN];
1028 printf("icmp_send dst %s src %s\n",
1029 inet_ntoa_r(ip->ip_dst, dstbuf),
1030 inet_ntoa_r(ip->ip_src, srcbuf));
1033 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
1037 * Return milliseconds since 00:00 UTC in network format.
1046 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1051 * Return the next larger or smaller MTU plateau (table from RFC 1191)
1052 * given current value MTU. If DIR is less than zero, a larger plateau
1053 * is returned; otherwise, a smaller value is returned.
1056 ip_next_mtu(int mtu, int dir)
1058 static int mtutab[] = {
1059 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
1064 size = (sizeof mtutab) / (sizeof mtutab[0]);
1066 for (i = 0; i < size; i++)
1067 if (mtu > mtutab[i])
1070 for (i = size - 1; i >= 0; i--)
1071 if (mtu < mtutab[i])
1073 if (mtu == mtutab[0])
1081 * badport_bandlim() - check for ICMP bandwidth limit
1083 * Return 0 if it is ok to send an ICMP error response, -1 if we have
1084 * hit our bandwidth limit and it is not ok.
1086 * If icmplim is <= 0, the feature is disabled and 0 is returned.
1088 * For now we separate the TCP and UDP subsystems w/ different 'which'
1089 * values. We may eventually remove this separation (and simplify the
1092 * Note that the printing of the error message is delayed so we can
1093 * properly print the icmp error rate that the system was trying to do
1094 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
1095 * the 'final' error, but it doesn't make sense to solve the printing
1096 * delay with more complex code.
1100 struct counter_rate cr;
1102 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
1103 { "icmp unreach response" },
1104 { "icmp ping response" },
1105 { "icmp tstamp response" },
1106 { "closed port RST response" },
1107 { "open port RST response" },
1108 { "icmp6 unreach response" },
1109 { "sctp ootb response" }
1111 #define V_icmp_rates VNET(icmp_rates)
1114 icmp_bandlimit_init(void)
1117 for (int i = 0; i < BANDLIM_MAX; i++) {
1118 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1119 V_icmp_rates[i].cr.cr_ticks = ticks;
1122 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1123 icmp_bandlimit_init, NULL);
1126 icmp_bandlimit_uninit(void)
1129 for (int i = 0; i < BANDLIM_MAX; i++)
1130 counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1132 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1133 icmp_bandlimit_uninit, NULL);
1136 badport_bandlim(int which)
1140 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1143 KASSERT(which >= 0 && which < BANDLIM_MAX,
1144 ("%s: which %d", __func__, which));
1146 if ((V_icmplim + V_icmplim_curr_jitter) <= 0)
1147 V_icmplim_curr_jitter = -V_icmplim + 1;
1149 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim +
1150 V_icmplim_curr_jitter);
1153 * Adjust limit +/- to jitter the measurement to deny a
1154 * side-channel port scan as in CVE-2020-25705
1156 if (V_icmplim_jitter > 0) {
1158 arc4random_uniform(V_icmplim_jitter * 2 +1)
1161 V_icmplim_curr_jitter = inc;
1166 if (pps > 0 && V_icmplim_output)
1167 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1168 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim +
1169 V_icmplim_curr_jitter);