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>
37 #include <sys/param.h>
38 #include <sys/systm.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
43 #include <sys/kernel.h>
45 #include <sys/rmlock.h>
46 #include <sys/sysctl.h>
47 #include <sys/syslog.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>
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>
74 #include <machine/in_cksum.h>
76 #include <security/mac/mac_framework.h>
80 * ICMP routines: error generation, receive packet processing, and
81 * routines to turnaround packets back to the originator, and
82 * host table maintenance routines.
84 VNET_DEFINE_STATIC(int, icmplim) = 200;
85 #define V_icmplim VNET(icmplim)
86 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
87 &VNET_NAME(icmplim), 0,
88 "Maximum number of ICMP responses per second");
90 VNET_DEFINE_STATIC(int, icmplim_curr_jitter) = 0;
91 #define V_icmplim_curr_jitter VNET(icmplim_curr_jitter)
92 VNET_DEFINE_STATIC(int, icmplim_jitter) = 16;
93 #define V_icmplim_jitter VNET(icmplim_jitter)
94 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLFLAG_VNET | CTLFLAG_RW,
95 &VNET_NAME(icmplim_jitter), 0,
96 "Random icmplim jitter adjustment limit");
98 VNET_DEFINE_STATIC(int, icmplim_output) = 1;
99 #define V_icmplim_output VNET(icmplim_output)
100 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
101 &VNET_NAME(icmplim_output), 0,
102 "Enable logging of ICMP response rate limiting");
105 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
106 VNET_PCPUSTAT_SYSINIT(icmpstat);
107 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
108 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
111 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
114 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
115 #define V_icmpmaskrepl VNET(icmpmaskrepl)
116 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
117 &VNET_NAME(icmpmaskrepl), 0,
118 "Reply to ICMP Address Mask Request packets");
120 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
121 #define V_icmpmaskfake VNET(icmpmaskfake)
122 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
123 &VNET_NAME(icmpmaskfake), 0,
124 "Fake reply to ICMP Address Mask Request packets");
126 VNET_DEFINE(int, drop_redirect) = 0;
127 #define V_drop_redirect VNET(drop_redirect)
128 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
129 &VNET_NAME(drop_redirect), 0,
130 "Ignore ICMP redirects");
132 VNET_DEFINE_STATIC(int, log_redirect) = 0;
133 #define V_log_redirect VNET(log_redirect)
134 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
135 &VNET_NAME(log_redirect), 0,
136 "Log ICMP redirects to the console");
138 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */
139 #define V_redirtimeout VNET(redirtimeout)
140 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW,
141 &VNET_NAME(redirtimeout), 0,
142 "Delay in seconds before expiring redirect route");
144 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
145 #define V_reply_src VNET(reply_src)
146 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
147 &VNET_NAME(reply_src), IFNAMSIZ,
148 "ICMP reply source for non-local packets");
150 VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
151 #define V_icmp_rfi VNET(icmp_rfi)
152 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
153 &VNET_NAME(icmp_rfi), 0,
154 "ICMP reply from incoming interface for non-local packets");
155 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
156 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
157 #define V_icmp_quotelen VNET(icmp_quotelen)
158 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
159 &VNET_NAME(icmp_quotelen), 0,
160 "Number of bytes from original packet to quote in ICMP reply");
162 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
163 #define V_icmpbmcastecho VNET(icmpbmcastecho)
164 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
165 &VNET_NAME(icmpbmcastecho), 0,
166 "Reply to multicast ICMP Echo Request and Timestamp packets");
168 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
169 #define V_icmptstamprepl VNET(icmptstamprepl)
170 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
171 &VNET_NAME(icmptstamprepl), 0,
172 "Respond to ICMP Timestamp packets");
174 VNET_DEFINE_STATIC(int, error_keeptags) = 0;
175 #define V_error_keeptags VNET(error_keeptags)
176 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
177 &VNET_NAME(error_keeptags), 0,
178 "ICMP error response keeps copy of mbuf_tags of original packet");
184 static void icmp_reflect(struct mbuf *);
185 static void icmp_send(struct mbuf *, struct mbuf *);
186 static int icmp_verify_redirect_gateway(struct sockaddr_in *,
187 struct sockaddr_in *, struct sockaddr_in *, u_int);
189 extern struct protosw inetsw[];
192 * Kernel module interface for updating icmpstat. The argument is an index
193 * into icmpstat treated as an array of u_long. While this encodes the
194 * general layout of icmpstat into the caller, it doesn't encode its
195 * location, so that future changes to add, for example, per-CPU stats
196 * support won't cause binary compatibility problems for kernel modules.
199 kmod_icmpstat_inc(int statnum)
202 counter_u64_add(VNET(icmpstat)[statnum], 1);
206 * Generate an error packet of type error
207 * in response to bad packet ip.
210 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
212 struct ip *oip, *nip;
215 unsigned icmplen, icmpelen, nlen, oiphlen;
217 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
220 if (type != ICMP_REDIRECT)
221 ICMPSTAT_INC(icps_error);
224 * if the original packet was encrypted.
225 * if not the first fragment of message.
226 * in response to a multicast or broadcast packet.
227 * if the old packet protocol was an ICMP error message.
229 if (n->m_flags & M_DECRYPTED)
231 if (n->m_flags & (M_BCAST|M_MCAST))
234 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
235 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
237 oip = mtod(n, struct ip *);
238 oiphlen = oip->ip_hl << 2;
239 if (n->m_len < oiphlen + ICMP_MINLEN)
243 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
245 if (oip->ip_off & htons(~(IP_MF|IP_DF)))
247 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
248 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
249 oiphlen))->icmp_type)) {
250 ICMPSTAT_INC(icps_oldicmp);
254 * Calculate length to quote from original packet and
255 * prevent the ICMP mbuf from overflowing.
256 * Unfortunately this is non-trivial since ip_forward()
257 * sends us truncated packets.
259 nlen = m_length(n, NULL);
260 if (oip->ip_p == IPPROTO_TCP) {
264 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
267 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
268 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
270 oip = mtod(n, struct ip *);
271 th = mtodo(n, oiphlen);
272 tcphlen = th->th_off << 2;
273 if (tcphlen < sizeof(struct tcphdr))
275 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
277 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
279 if (n->m_len < oiphlen + tcphlen &&
280 (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
282 oip = mtod(n, struct ip *);
283 icmpelen = max(tcphlen, min(V_icmp_quotelen,
284 ntohs(oip->ip_len) - oiphlen));
285 } else if (oip->ip_p == IPPROTO_SCTP) {
287 struct sctp_chunkhdr *ch;
289 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
291 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
294 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
295 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
297 oip = mtod(n, struct ip *);
298 icmpelen = max(sizeof(struct sctphdr),
299 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
300 sh = mtodo(n, oiphlen);
301 if (ntohl(sh->v_tag) == 0 &&
302 ntohs(oip->ip_len) >= oiphlen +
303 sizeof(struct sctphdr) + 8 &&
304 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
305 n->m_next != NULL)) {
306 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
307 (n = m_pullup(n, oiphlen +
308 sizeof(struct sctphdr) + 8)) == NULL)
310 oip = mtod(n, struct ip *);
311 sh = mtodo(n, oiphlen);
312 ch = (struct sctp_chunkhdr *)(sh + 1);
313 if (ch->chunk_type == SCTP_INITIATION) {
314 icmpelen = max(sizeof(struct sctphdr) + 8,
315 min(V_icmp_quotelen, ntohs(oip->ip_len) -
320 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
323 icmplen = min(oiphlen + icmpelen, nlen);
324 if (icmplen < sizeof(struct ip))
327 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
328 m = m_gethdr(M_NOWAIT, MT_DATA);
330 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
334 mac_netinet_icmp_reply(n, m);
336 icmplen = min(icmplen, M_TRAILINGSPACE(m) -
337 sizeof(struct ip) - ICMP_MINLEN);
338 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
339 m->m_data += sizeof(struct ip);
340 m->m_len = ICMP_MINLEN + icmplen;
342 /* XXX MRT make the outgoing packet use the same FIB
343 * that was associated with the incoming packet
345 M_SETFIB(m, M_GETFIB(n));
346 icp = mtod(m, struct icmp *);
347 ICMPSTAT_INC(icps_outhist[type]);
348 icp->icmp_type = type;
349 if (type == ICMP_REDIRECT)
350 icp->icmp_gwaddr.s_addr = dest;
354 * The following assignments assume an overlay with the
355 * just zeroed icmp_void field.
357 if (type == ICMP_PARAMPROB) {
358 icp->icmp_pptr = code;
360 } else if (type == ICMP_UNREACH &&
361 code == ICMP_UNREACH_NEEDFRAG && mtu) {
362 icp->icmp_nextmtu = htons(mtu);
365 icp->icmp_code = code;
368 * Copy the quotation into ICMP message and
369 * convert quoted IP header back to network representation.
371 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
375 * Set up ICMP message mbuf and copy old IP header (without options
376 * in front of ICMP message.
377 * If the original mbuf was meant to bypass the firewall, the error
378 * reply should bypass as well.
380 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
381 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
382 ("insufficient space for ip header"));
383 m->m_data -= sizeof(struct ip);
384 m->m_len += sizeof(struct ip);
385 m->m_pkthdr.len = m->m_len;
386 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
387 nip = mtod(m, struct ip *);
388 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
389 nip->ip_len = htons(m->m_len);
390 nip->ip_v = IPVERSION;
392 nip->ip_p = IPPROTO_ICMP;
396 if (V_error_keeptags)
397 m_tag_copy_chain(m, n, M_NOWAIT);
406 * Process a received ICMP message.
409 icmp_input(struct mbuf **mp, int *offp, int proto)
412 struct in_ifaddr *ia;
413 struct mbuf *m = *mp;
414 struct ip *ip = mtod(m, struct ip *);
415 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
417 int icmplen = ntohs(ip->ip_len) - *offp;
419 void (*ctlfunc)(int, struct sockaddr *, void *);
427 * Locate icmp structure in mbuf, and check
428 * that not corrupted and of at least minimum length.
432 char srcbuf[INET_ADDRSTRLEN];
433 char dstbuf[INET_ADDRSTRLEN];
435 printf("icmp_input from %s to %s, len %d\n",
436 inet_ntoa_r(ip->ip_src, srcbuf),
437 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
440 if (icmplen < ICMP_MINLEN) {
441 ICMPSTAT_INC(icps_tooshort);
444 i = hlen + min(icmplen, ICMP_ADVLENMIN);
445 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
446 ICMPSTAT_INC(icps_tooshort);
447 return (IPPROTO_DONE);
449 ip = mtod(m, struct ip *);
452 icp = mtod(m, struct icmp *);
453 if (in_cksum(m, icmplen)) {
454 ICMPSTAT_INC(icps_checksum);
462 printf("icmp_input, type %d code %d\n", icp->icmp_type,
467 * Message type specific processing.
469 if (icp->icmp_type > ICMP_MAXTYPE)
473 bzero(&icmpsrc, sizeof(icmpsrc));
474 icmpsrc.sin_len = sizeof(struct sockaddr_in);
475 icmpsrc.sin_family = AF_INET;
476 bzero(&icmpdst, sizeof(icmpdst));
477 icmpdst.sin_len = sizeof(struct sockaddr_in);
478 icmpdst.sin_family = AF_INET;
479 bzero(&icmpgw, sizeof(icmpgw));
480 icmpgw.sin_len = sizeof(struct sockaddr_in);
481 icmpgw.sin_family = AF_INET;
483 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
484 code = icp->icmp_code;
485 switch (icp->icmp_type) {
488 case ICMP_UNREACH_NET:
489 case ICMP_UNREACH_HOST:
490 case ICMP_UNREACH_SRCFAIL:
491 case ICMP_UNREACH_NET_UNKNOWN:
492 case ICMP_UNREACH_HOST_UNKNOWN:
493 case ICMP_UNREACH_ISOLATED:
494 case ICMP_UNREACH_TOSNET:
495 case ICMP_UNREACH_TOSHOST:
496 case ICMP_UNREACH_HOST_PRECEDENCE:
497 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
498 code = PRC_UNREACH_NET;
501 case ICMP_UNREACH_NEEDFRAG:
506 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
507 * Treat subcodes 2,3 as immediate RST
509 case ICMP_UNREACH_PROTOCOL:
510 code = PRC_UNREACH_PROTOCOL;
512 case ICMP_UNREACH_PORT:
513 code = PRC_UNREACH_PORT;
516 case ICMP_UNREACH_NET_PROHIB:
517 case ICMP_UNREACH_HOST_PROHIB:
518 case ICMP_UNREACH_FILTER_PROHIB:
519 code = PRC_UNREACH_ADMIN_PROHIB;
530 code += PRC_TIMXCEED_INTRANS;
536 code = PRC_PARAMPROB;
539 * Problem with datagram; advise higher level routines.
541 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
542 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
543 ICMPSTAT_INC(icps_badlen);
546 /* Discard ICMP's in response to multicast packets */
547 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
551 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
553 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
555 * XXX if the packet contains [IPv4 AH TCP], we can't make a
556 * notification to TCP layer.
558 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
560 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
561 /* This should actually not happen */
562 ICMPSTAT_INC(icps_tooshort);
563 return (IPPROTO_DONE);
565 ip = mtod(m, struct ip *);
566 icp = (struct icmp *)(ip + 1);
568 * The upper layer handler can rely on:
569 * - The outer IP header has no options.
570 * - The outer IP header, the ICMP header, the inner IP header,
571 * and the first n bytes of the inner payload are contiguous.
572 * n is at least 8, but might be larger based on
573 * ICMP_ADVLENPREF. See its definition in ip_icmp.h.
575 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
577 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
578 (void *)&icp->icmp_ip);
582 ICMPSTAT_INC(icps_badcode);
586 if (!V_icmpbmcastecho
587 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
588 ICMPSTAT_INC(icps_bmcastecho);
591 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
593 icp->icmp_type = ICMP_ECHOREPLY;
597 if (V_icmptstamprepl == 0)
599 if (!V_icmpbmcastecho
600 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
601 ICMPSTAT_INC(icps_bmcasttstamp);
604 if (icmplen < ICMP_TSLEN) {
605 ICMPSTAT_INC(icps_badlen);
608 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
610 icp->icmp_type = ICMP_TSTAMPREPLY;
611 icp->icmp_rtime = iptime();
612 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
616 if (V_icmpmaskrepl == 0)
619 * We are not able to respond with all ones broadcast
620 * unless we receive it over a point-to-point interface.
622 if (icmplen < ICMP_MASKLEN)
624 switch (ip->ip_dst.s_addr) {
625 case INADDR_BROADCAST:
627 icmpdst.sin_addr = ip->ip_src;
631 icmpdst.sin_addr = ip->ip_dst;
633 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
634 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
637 if (ia->ia_ifp == NULL)
639 icp->icmp_type = ICMP_MASKREPLY;
640 if (V_icmpmaskfake == 0)
641 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
643 icp->icmp_mask = V_icmpmaskfake;
644 if (ip->ip_src.s_addr == 0) {
645 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
646 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
647 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
648 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
651 ICMPSTAT_INC(icps_reflect);
652 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
654 return (IPPROTO_DONE);
657 if (V_log_redirect) {
660 src = ntohl(ip->ip_src.s_addr);
661 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
662 gw = ntohl(icp->icmp_gwaddr.s_addr);
663 printf("icmp redirect from %d.%d.%d.%d: "
664 "%d.%d.%d.%d => %d.%d.%d.%d\n",
665 (int)(src >> 24), (int)((src >> 16) & 0xff),
666 (int)((src >> 8) & 0xff), (int)(src & 0xff),
667 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
668 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
669 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
670 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
673 * RFC1812 says we must ignore ICMP redirects if we
674 * are acting as router.
676 if (V_drop_redirect || V_ipforwarding)
680 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
681 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
682 ICMPSTAT_INC(icps_badlen);
686 * Short circuit routing redirects to force
687 * immediate change in the kernel's routing
688 * tables. The message is also handed to anyone
689 * listening on a raw socket (e.g. the routing
690 * daemon for use in updating its tables).
692 icmpgw.sin_addr = ip->ip_src;
693 icmpdst.sin_addr = icp->icmp_gwaddr;
696 char dstbuf[INET_ADDRSTRLEN];
697 char gwbuf[INET_ADDRSTRLEN];
699 printf("redirect dst %s to %s\n",
700 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
701 inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
704 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
707 * RFC 1122 says network (code 0,2) redirects SHOULD
708 * be treated identically to the host redirects.
709 * Given that, ignore network masks.
714 * icmpsrc: route destination
715 * icmpdst: route gateway
716 * icmpgw: message source
719 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst,
721 /* TODO: increment bad redirects here */
725 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
726 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc,
727 (struct sockaddr *)&icmpdst,
728 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif,
729 RTF_GATEWAY, V_redirtimeout);
731 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
735 * No kernel processing for the following;
736 * just fall through to send to raw listener.
739 case ICMP_ROUTERADVERT:
740 case ICMP_ROUTERSOLICIT:
741 case ICMP_TSTAMPREPLY:
744 case ICMP_SOURCEQUENCH:
751 rip_input(mp, offp, proto);
752 return (IPPROTO_DONE);
756 return (IPPROTO_DONE);
760 * Reflect the ip packet back to the source
763 icmp_reflect(struct mbuf *m)
765 struct rm_priotracker in_ifa_tracker;
766 struct ip *ip = mtod(m, struct ip *);
769 struct in_ifaddr *ia;
771 struct nhop_object *nh;
772 struct mbuf *opts = NULL;
773 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
777 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
778 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
779 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
780 m_freem(m); /* Bad return address */
781 ICMPSTAT_INC(icps_badaddr);
782 goto done; /* Ip_output() will check for broadcast */
786 ip->ip_dst = ip->ip_src;
789 * Source selection for ICMP replies:
791 * If the incoming packet was addressed directly to one of our
792 * own addresses, use dst as the src for the reply.
794 IN_IFADDR_RLOCK(&in_ifa_tracker);
795 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
796 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
797 t = IA_SIN(ia)->sin_addr;
798 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
802 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
805 * If the incoming packet was addressed to one of our broadcast
806 * addresses, use the first non-broadcast address which corresponds
807 * to the incoming interface.
809 ifp = m->m_pkthdr.rcvif;
810 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
811 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
812 if (ifa->ifa_addr->sa_family != AF_INET)
815 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
817 t = IA_SIN(ia)->sin_addr;
823 * If the packet was transiting through us, use the address of
824 * the interface the packet came through in. If that interface
825 * doesn't have a suitable IP address, the normal selection
828 if (V_icmp_rfi && ifp != NULL) {
829 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
830 if (ifa->ifa_addr->sa_family != AF_INET)
833 t = IA_SIN(ia)->sin_addr;
838 * If the incoming packet was not addressed directly to us, use
839 * designated interface for icmp replies specified by sysctl
840 * net.inet.icmp.reply_src (default not set). Otherwise continue
841 * with normal source selection.
843 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
844 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
845 if (ifa->ifa_addr->sa_family != AF_INET)
848 t = IA_SIN(ia)->sin_addr;
853 * If the packet was transiting through us, use the address of
854 * the interface that is the closest to the packet source.
855 * When we don't have a route back to the packet source, stop here
856 * and drop the packet.
858 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
861 ICMPSTAT_INC(icps_noroute);
864 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr;
867 mac_netinet_icmp_replyinplace(m);
870 ip->ip_ttl = V_ip_defttl;
878 * Retrieve any source routing from the incoming packet;
879 * add on any record-route or timestamp options.
881 cp = (u_char *) (ip + 1);
882 if ((opts = ip_srcroute(m)) == NULL &&
883 (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
884 opts->m_len = sizeof(struct in_addr);
885 mtod(opts, struct in_addr *)->s_addr = 0;
890 printf("icmp_reflect optlen %d rt %d => ",
891 optlen, opts->m_len);
893 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
894 opt = cp[IPOPT_OPTVAL];
895 if (opt == IPOPT_EOL)
897 if (opt == IPOPT_NOP)
900 if (cnt < IPOPT_OLEN + sizeof(*cp))
902 len = cp[IPOPT_OLEN];
903 if (len < IPOPT_OLEN + sizeof(*cp) ||
908 * Should check for overflow, but it "can't happen"
910 if (opt == IPOPT_RR || opt == IPOPT_TS ||
911 opt == IPOPT_SECURITY) {
913 mtod(opts, caddr_t) + opts->m_len, len);
917 /* Terminate & pad, if necessary */
918 cnt = opts->m_len % 4;
920 for (; cnt < 4; cnt++) {
921 *(mtod(opts, caddr_t) + opts->m_len) =
928 printf("%d\n", opts->m_len);
933 m_tag_delete_nonpersistent(m);
934 m->m_flags &= ~(M_BCAST|M_MCAST);
942 * Verifies if redirect message is valid, according to RFC 1122
944 * @src: sockaddr with address of redirect originator
945 * @dst: sockaddr with destination in question
946 * @gateway: new proposed gateway
948 * Returns 0 on success.
951 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst,
952 struct sockaddr_in *gateway, u_int fibnum)
954 struct nhop_object *nh;
959 /* Verify the gateway is directly reachable. */
960 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL)
961 return (ENETUNREACH);
963 /* TODO: fib-aware. */
964 if (ifa_ifwithaddr_check((struct sockaddr *)gateway))
965 return (EHOSTUNREACH);
967 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0);
972 * If the redirect isn't from our current router for this dst,
973 * it's either old or wrong. If it redirects us to ourselves,
974 * we have a routing loop, perhaps as a result of an interface
975 * going down recently.
977 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa))
979 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK)
982 /* If host route already exists, ignore redirect. */
983 if (nh->nh_flags & NHF_HOST)
986 /* If the prefix is directly reachable, ignore redirect. */
987 if (!(nh->nh_flags & NHF_GATEWAY))
994 * Send an icmp packet back to the ip level,
995 * after supplying a checksum.
998 icmp_send(struct mbuf *m, struct mbuf *opts)
1000 struct ip *ip = mtod(m, struct ip *);
1004 hlen = ip->ip_hl << 2;
1007 icp = mtod(m, struct icmp *);
1008 icp->icmp_cksum = 0;
1009 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
1012 m->m_pkthdr.rcvif = (struct ifnet *)0;
1015 char dstbuf[INET_ADDRSTRLEN];
1016 char srcbuf[INET_ADDRSTRLEN];
1018 printf("icmp_send dst %s src %s\n",
1019 inet_ntoa_r(ip->ip_dst, dstbuf),
1020 inet_ntoa_r(ip->ip_src, srcbuf));
1023 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
1027 * Return milliseconds since 00:00 UTC in network format.
1036 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1041 * Return the next larger or smaller MTU plateau (table from RFC 1191)
1042 * given current value MTU. If DIR is less than zero, a larger plateau
1043 * is returned; otherwise, a smaller value is returned.
1046 ip_next_mtu(int mtu, int dir)
1048 static int mtutab[] = {
1049 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
1054 size = (sizeof mtutab) / (sizeof mtutab[0]);
1056 for (i = 0; i < size; i++)
1057 if (mtu > mtutab[i])
1060 for (i = size - 1; i >= 0; i--)
1061 if (mtu < mtutab[i])
1063 if (mtu == mtutab[0])
1071 * badport_bandlim() - check for ICMP bandwidth limit
1073 * Return 0 if it is ok to send an ICMP error response, -1 if we have
1074 * hit our bandwidth limit and it is not ok.
1076 * If icmplim is <= 0, the feature is disabled and 0 is returned.
1078 * For now we separate the TCP and UDP subsystems w/ different 'which'
1079 * values. We may eventually remove this separation (and simplify the
1082 * Note that the printing of the error message is delayed so we can
1083 * properly print the icmp error rate that the system was trying to do
1084 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
1085 * the 'final' error, but it doesn't make sense to solve the printing
1086 * delay with more complex code.
1090 struct counter_rate cr;
1092 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
1093 { "icmp unreach response" },
1094 { "icmp ping response" },
1095 { "icmp tstamp response" },
1096 { "closed port RST response" },
1097 { "open port RST response" },
1098 { "icmp6 unreach response" },
1099 { "sctp ootb response" }
1101 #define V_icmp_rates VNET(icmp_rates)
1104 icmp_bandlimit_init(void)
1107 for (int i = 0; i < BANDLIM_MAX; i++) {
1108 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1109 V_icmp_rates[i].cr.cr_ticks = ticks;
1112 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1113 icmp_bandlimit_init, NULL);
1116 icmp_bandlimit_uninit(void)
1119 for (int i = 0; i < BANDLIM_MAX; i++)
1120 counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1122 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1123 icmp_bandlimit_uninit, NULL);
1126 badport_bandlim(int which)
1130 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1133 KASSERT(which >= 0 && which < BANDLIM_MAX,
1134 ("%s: which %d", __func__, which));
1136 if ((V_icmplim + V_icmplim_curr_jitter) <= 0)
1137 V_icmplim_curr_jitter = -V_icmplim + 1;
1139 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim +
1140 V_icmplim_curr_jitter);
1143 * Adjust limit +/- to jitter the measurement to deny a
1144 * side-channel port scan as in CVE-2020-25705
1146 if (V_icmplim_jitter > 0) {
1148 arc4random_uniform(V_icmplim_jitter * 2 +1)
1151 V_icmplim_curr_jitter = inc;
1156 if (pps > 0 && V_icmplim_output)
1157 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1158 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim +
1159 V_icmplim_curr_jitter);