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/rmlock.h>
48 #include <sys/sysctl.h>
49 #include <sys/syslog.h>
52 #include <net/if_var.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/tcp_var.h>
71 #include <netinet/tcpip.h>
72 #include <netinet/icmp_var.h>
76 #include <machine/in_cksum.h>
78 #include <security/mac/mac_framework.h>
82 * ICMP routines: error generation, receive packet processing, and
83 * routines to turnaround packets back to the originator, and
84 * host table maintenance routines.
86 VNET_DEFINE_STATIC(int, icmplim) = 200;
87 #define V_icmplim VNET(icmplim)
88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
89 &VNET_NAME(icmplim), 0,
90 "Maximum number of ICMP responses per second");
92 VNET_DEFINE_STATIC(int, icmplim_curr_jitter) = 0;
93 #define V_icmplim_curr_jitter VNET(icmplim_curr_jitter)
94 VNET_DEFINE_STATIC(int, icmplim_jitter) = 16;
95 #define V_icmplim_jitter VNET(icmplim_jitter)
96 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLFLAG_VNET | CTLFLAG_RW,
97 &VNET_NAME(icmplim_jitter), 0,
98 "Random icmplim jitter adjustment limit");
100 VNET_DEFINE_STATIC(int, icmplim_output) = 1;
101 #define V_icmplim_output VNET(icmplim_output)
102 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
103 &VNET_NAME(icmplim_output), 0,
104 "Enable logging of ICMP response rate limiting");
107 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
108 VNET_PCPUSTAT_SYSINIT(icmpstat);
109 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
110 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
113 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
116 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
117 #define V_icmpmaskrepl VNET(icmpmaskrepl)
118 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
119 &VNET_NAME(icmpmaskrepl), 0,
120 "Reply to ICMP Address Mask Request packets");
122 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
123 #define V_icmpmaskfake VNET(icmpmaskfake)
124 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
125 &VNET_NAME(icmpmaskfake), 0,
126 "Fake reply to ICMP Address Mask Request packets");
128 VNET_DEFINE(int, drop_redirect) = 0;
129 #define V_drop_redirect VNET(drop_redirect)
130 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
131 &VNET_NAME(drop_redirect), 0,
132 "Ignore ICMP redirects");
134 VNET_DEFINE_STATIC(int, log_redirect) = 0;
135 #define V_log_redirect VNET(log_redirect)
136 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
137 &VNET_NAME(log_redirect), 0,
138 "Log ICMP redirects to the console");
140 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */
141 #define V_redirtimeout VNET(redirtimeout)
142 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW,
143 &VNET_NAME(redirtimeout), 0,
144 "Delay in seconds before expiring redirect route");
146 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
147 #define V_reply_src VNET(reply_src)
148 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
149 &VNET_NAME(reply_src), IFNAMSIZ,
150 "ICMP reply source for non-local packets");
152 VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
153 #define V_icmp_rfi VNET(icmp_rfi)
154 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
155 &VNET_NAME(icmp_rfi), 0,
156 "ICMP reply from incoming interface for non-local packets");
157 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
158 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
159 #define V_icmp_quotelen VNET(icmp_quotelen)
160 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
161 &VNET_NAME(icmp_quotelen), 0,
162 "Number of bytes from original packet to quote in ICMP reply");
164 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
165 #define V_icmpbmcastecho VNET(icmpbmcastecho)
166 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
167 &VNET_NAME(icmpbmcastecho), 0,
168 "Reply to multicast ICMP Echo Request and Timestamp packets");
170 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
171 #define V_icmptstamprepl VNET(icmptstamprepl)
172 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
173 &VNET_NAME(icmptstamprepl), 0,
174 "Respond to ICMP Timestamp packets");
176 VNET_DEFINE_STATIC(int, error_keeptags) = 0;
177 #define V_error_keeptags VNET(error_keeptags)
178 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
179 &VNET_NAME(error_keeptags), 0,
180 "ICMP error response keeps copy of mbuf_tags of original packet");
186 static void icmp_reflect(struct mbuf *);
187 static void icmp_send(struct mbuf *, struct mbuf *);
188 static int icmp_verify_redirect_gateway(struct sockaddr_in *,
189 struct sockaddr_in *, struct sockaddr_in *, u_int);
191 extern struct protosw inetsw[];
194 * Kernel module interface for updating icmpstat. The argument is an index
195 * into icmpstat treated as an array of u_long. While this encodes the
196 * general layout of icmpstat into the caller, it doesn't encode its
197 * location, so that future changes to add, for example, per-CPU stats
198 * support won't cause binary compatibility problems for kernel modules.
201 kmod_icmpstat_inc(int statnum)
204 counter_u64_add(VNET(icmpstat)[statnum], 1);
208 * Generate an error packet of type error
209 * in response to bad packet ip.
212 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
214 struct ip *oip, *nip;
217 unsigned icmplen, icmpelen, nlen, oiphlen;
219 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
222 if (type != ICMP_REDIRECT)
223 ICMPSTAT_INC(icps_error);
226 * if the original packet was encrypted.
227 * if not the first fragment of message.
228 * in response to a multicast or broadcast packet.
229 * if the old packet protocol was an ICMP error message.
231 if (n->m_flags & M_DECRYPTED)
233 if (n->m_flags & (M_BCAST|M_MCAST))
236 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
237 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
239 oip = mtod(n, struct ip *);
240 oiphlen = oip->ip_hl << 2;
241 if (n->m_len < oiphlen + ICMP_MINLEN)
245 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
247 if (oip->ip_off & htons(~(IP_MF|IP_DF)))
249 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
250 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
251 oiphlen))->icmp_type)) {
252 ICMPSTAT_INC(icps_oldicmp);
256 * Calculate length to quote from original packet and
257 * prevent the ICMP mbuf from overflowing.
258 * Unfortunately this is non-trivial since ip_forward()
259 * sends us truncated packets.
261 nlen = m_length(n, NULL);
262 if (oip->ip_p == IPPROTO_TCP) {
266 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
269 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
270 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
272 oip = mtod(n, struct ip *);
273 th = mtodo(n, oiphlen);
274 tcphlen = th->th_off << 2;
275 if (tcphlen < sizeof(struct tcphdr))
277 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
279 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
281 if (n->m_len < oiphlen + tcphlen &&
282 (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
284 oip = mtod(n, struct ip *);
285 icmpelen = max(tcphlen, min(V_icmp_quotelen,
286 ntohs(oip->ip_len) - oiphlen));
287 } else if (oip->ip_p == IPPROTO_SCTP) {
289 struct sctp_chunkhdr *ch;
291 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
293 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
296 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
297 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
299 oip = mtod(n, struct ip *);
300 icmpelen = max(sizeof(struct sctphdr),
301 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
302 sh = mtodo(n, oiphlen);
303 if (ntohl(sh->v_tag) == 0 &&
304 ntohs(oip->ip_len) >= oiphlen +
305 sizeof(struct sctphdr) + 8 &&
306 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
307 n->m_next != NULL)) {
308 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
309 (n = m_pullup(n, oiphlen +
310 sizeof(struct sctphdr) + 8)) == NULL)
312 oip = mtod(n, struct ip *);
313 sh = mtodo(n, oiphlen);
314 ch = (struct sctp_chunkhdr *)(sh + 1);
315 if (ch->chunk_type == SCTP_INITIATION) {
316 icmpelen = max(sizeof(struct sctphdr) + 8,
317 min(V_icmp_quotelen, ntohs(oip->ip_len) -
322 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
325 icmplen = min(oiphlen + icmpelen, nlen);
326 if (icmplen < sizeof(struct ip))
329 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
330 m = m_gethdr(M_NOWAIT, MT_DATA);
332 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
336 mac_netinet_icmp_reply(n, m);
338 icmplen = min(icmplen, M_TRAILINGSPACE(m) -
339 sizeof(struct ip) - ICMP_MINLEN);
340 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
341 m->m_data += sizeof(struct ip);
342 m->m_len = ICMP_MINLEN + icmplen;
344 /* XXX MRT make the outgoing packet use the same FIB
345 * that was associated with the incoming packet
347 M_SETFIB(m, M_GETFIB(n));
348 icp = mtod(m, struct icmp *);
349 ICMPSTAT_INC(icps_outhist[type]);
350 icp->icmp_type = type;
351 if (type == ICMP_REDIRECT)
352 icp->icmp_gwaddr.s_addr = dest;
356 * The following assignments assume an overlay with the
357 * just zeroed icmp_void field.
359 if (type == ICMP_PARAMPROB) {
360 icp->icmp_pptr = code;
362 } else if (type == ICMP_UNREACH &&
363 code == ICMP_UNREACH_NEEDFRAG && mtu) {
364 icp->icmp_nextmtu = htons(mtu);
367 icp->icmp_code = code;
370 * Copy the quotation into ICMP message and
371 * convert quoted IP header back to network representation.
373 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
377 * Set up ICMP message mbuf and copy old IP header (without options
378 * in front of ICMP message.
379 * If the original mbuf was meant to bypass the firewall, the error
380 * reply should bypass as well.
382 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
383 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
384 ("insufficient space for ip header"));
385 m->m_data -= sizeof(struct ip);
386 m->m_len += sizeof(struct ip);
387 m->m_pkthdr.len = m->m_len;
388 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
389 nip = mtod(m, struct ip *);
390 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
391 nip->ip_len = htons(m->m_len);
392 nip->ip_v = IPVERSION;
394 nip->ip_p = IPPROTO_ICMP;
398 if (V_error_keeptags)
399 m_tag_copy_chain(m, n, M_NOWAIT);
408 * Process a received ICMP message.
411 icmp_input(struct mbuf **mp, int *offp, int proto)
414 struct in_ifaddr *ia;
415 struct mbuf *m = *mp;
416 struct ip *ip = mtod(m, struct ip *);
417 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
419 int icmplen = ntohs(ip->ip_len) - *offp;
421 void (*ctlfunc)(int, struct sockaddr *, void *);
429 * Locate icmp structure in mbuf, and check
430 * that not corrupted and of at least minimum length.
434 char srcbuf[INET_ADDRSTRLEN];
435 char dstbuf[INET_ADDRSTRLEN];
437 printf("icmp_input from %s to %s, len %d\n",
438 inet_ntoa_r(ip->ip_src, srcbuf),
439 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
442 if (icmplen < ICMP_MINLEN) {
443 ICMPSTAT_INC(icps_tooshort);
446 i = hlen + min(icmplen, ICMP_ADVLENMIN);
447 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
448 ICMPSTAT_INC(icps_tooshort);
449 return (IPPROTO_DONE);
451 ip = mtod(m, struct ip *);
454 icp = mtod(m, struct icmp *);
455 if (in_cksum(m, icmplen)) {
456 ICMPSTAT_INC(icps_checksum);
464 printf("icmp_input, type %d code %d\n", icp->icmp_type,
469 * Message type specific processing.
471 if (icp->icmp_type > ICMP_MAXTYPE)
475 bzero(&icmpsrc, sizeof(icmpsrc));
476 icmpsrc.sin_len = sizeof(struct sockaddr_in);
477 icmpsrc.sin_family = AF_INET;
478 bzero(&icmpdst, sizeof(icmpdst));
479 icmpdst.sin_len = sizeof(struct sockaddr_in);
480 icmpdst.sin_family = AF_INET;
481 bzero(&icmpgw, sizeof(icmpgw));
482 icmpgw.sin_len = sizeof(struct sockaddr_in);
483 icmpgw.sin_family = AF_INET;
485 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
486 code = icp->icmp_code;
487 switch (icp->icmp_type) {
490 case ICMP_UNREACH_NET:
491 case ICMP_UNREACH_HOST:
492 case ICMP_UNREACH_SRCFAIL:
493 case ICMP_UNREACH_NET_UNKNOWN:
494 case ICMP_UNREACH_HOST_UNKNOWN:
495 case ICMP_UNREACH_ISOLATED:
496 case ICMP_UNREACH_TOSNET:
497 case ICMP_UNREACH_TOSHOST:
498 case ICMP_UNREACH_HOST_PRECEDENCE:
499 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
500 code = PRC_UNREACH_NET;
503 case ICMP_UNREACH_NEEDFRAG:
508 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
509 * Treat subcodes 2,3 as immediate RST
511 case ICMP_UNREACH_PROTOCOL:
512 code = PRC_UNREACH_PROTOCOL;
514 case ICMP_UNREACH_PORT:
515 code = PRC_UNREACH_PORT;
518 case ICMP_UNREACH_NET_PROHIB:
519 case ICMP_UNREACH_HOST_PROHIB:
520 case ICMP_UNREACH_FILTER_PROHIB:
521 code = PRC_UNREACH_ADMIN_PROHIB;
532 code += PRC_TIMXCEED_INTRANS;
538 code = PRC_PARAMPROB;
541 * Problem with datagram; advise higher level routines.
543 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
544 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
545 ICMPSTAT_INC(icps_badlen);
548 /* Discard ICMP's in response to multicast packets */
549 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
553 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
555 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
557 * XXX if the packet contains [IPv4 AH TCP], we can't make a
558 * notification to TCP layer.
560 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
562 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
563 /* This should actually not happen */
564 ICMPSTAT_INC(icps_tooshort);
565 return (IPPROTO_DONE);
567 ip = mtod(m, struct ip *);
568 icp = (struct icmp *)(ip + 1);
570 * The upper layer handler can rely on:
571 * - The outer IP header has no options.
572 * - The outer IP header, the ICMP header, the inner IP header,
573 * and the first n bytes of the inner payload are contiguous.
574 * n is at least 8, but might be larger based on
575 * ICMP_ADVLENPREF. See its definition in ip_icmp.h.
577 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
579 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
580 (void *)&icp->icmp_ip);
584 ICMPSTAT_INC(icps_badcode);
588 if (!V_icmpbmcastecho
589 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
590 ICMPSTAT_INC(icps_bmcastecho);
593 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
595 icp->icmp_type = ICMP_ECHOREPLY;
599 if (V_icmptstamprepl == 0)
601 if (!V_icmpbmcastecho
602 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
603 ICMPSTAT_INC(icps_bmcasttstamp);
606 if (icmplen < ICMP_TSLEN) {
607 ICMPSTAT_INC(icps_badlen);
610 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
612 icp->icmp_type = ICMP_TSTAMPREPLY;
613 icp->icmp_rtime = iptime();
614 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
618 if (V_icmpmaskrepl == 0)
621 * We are not able to respond with all ones broadcast
622 * unless we receive it over a point-to-point interface.
624 if (icmplen < ICMP_MASKLEN)
626 switch (ip->ip_dst.s_addr) {
627 case INADDR_BROADCAST:
629 icmpdst.sin_addr = ip->ip_src;
633 icmpdst.sin_addr = ip->ip_dst;
635 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
636 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
639 if (ia->ia_ifp == NULL)
641 icp->icmp_type = ICMP_MASKREPLY;
642 if (V_icmpmaskfake == 0)
643 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
645 icp->icmp_mask = V_icmpmaskfake;
646 if (ip->ip_src.s_addr == 0) {
647 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
648 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
649 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
650 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
653 ICMPSTAT_INC(icps_reflect);
654 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
656 return (IPPROTO_DONE);
659 if (V_log_redirect) {
662 src = ntohl(ip->ip_src.s_addr);
663 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
664 gw = ntohl(icp->icmp_gwaddr.s_addr);
665 printf("icmp redirect from %d.%d.%d.%d: "
666 "%d.%d.%d.%d => %d.%d.%d.%d\n",
667 (int)(src >> 24), (int)((src >> 16) & 0xff),
668 (int)((src >> 8) & 0xff), (int)(src & 0xff),
669 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
670 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
671 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
672 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
675 * RFC1812 says we must ignore ICMP redirects if we
676 * are acting as router.
678 if (V_drop_redirect || V_ipforwarding)
682 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
683 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
684 ICMPSTAT_INC(icps_badlen);
688 * Short circuit routing redirects to force
689 * immediate change in the kernel's routing
690 * tables. The message is also handed to anyone
691 * listening on a raw socket (e.g. the routing
692 * daemon for use in updating its tables).
694 icmpgw.sin_addr = ip->ip_src;
695 icmpdst.sin_addr = icp->icmp_gwaddr;
698 char dstbuf[INET_ADDRSTRLEN];
699 char gwbuf[INET_ADDRSTRLEN];
701 printf("redirect dst %s to %s\n",
702 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
703 inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
706 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
709 * RFC 1122 says network (code 0,2) redirects SHOULD
710 * be treated identically to the host redirects.
711 * Given that, ignore network masks.
716 * icmpsrc: route destination
717 * icmpdst: route gateway
718 * icmpgw: message source
721 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst,
723 /* TODO: increment bad redirects here */
727 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
728 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc,
729 (struct sockaddr *)&icmpdst,
730 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif,
731 RTF_GATEWAY, V_redirtimeout);
733 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
737 * No kernel processing for the following;
738 * just fall through to send to raw listener.
741 case ICMP_ROUTERADVERT:
742 case ICMP_ROUTERSOLICIT:
743 case ICMP_TSTAMPREPLY:
746 case ICMP_SOURCEQUENCH:
753 rip_input(mp, offp, proto);
754 return (IPPROTO_DONE);
758 return (IPPROTO_DONE);
762 * Reflect the ip packet back to the source
765 icmp_reflect(struct mbuf *m)
767 struct rm_priotracker in_ifa_tracker;
768 struct ip *ip = mtod(m, struct ip *);
771 struct in_ifaddr *ia;
773 struct nhop_object *nh;
774 struct mbuf *opts = NULL;
775 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
779 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
780 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
781 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
782 m_freem(m); /* Bad return address */
783 ICMPSTAT_INC(icps_badaddr);
784 goto done; /* Ip_output() will check for broadcast */
788 ip->ip_dst = ip->ip_src;
791 * Source selection for ICMP replies:
793 * If the incoming packet was addressed directly to one of our
794 * own addresses, use dst as the src for the reply.
796 IN_IFADDR_RLOCK(&in_ifa_tracker);
797 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
798 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
799 t = IA_SIN(ia)->sin_addr;
800 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
804 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
807 * If the incoming packet was addressed to one of our broadcast
808 * addresses, use the first non-broadcast address which corresponds
809 * to the incoming interface.
811 ifp = m->m_pkthdr.rcvif;
812 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
813 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
814 if (ifa->ifa_addr->sa_family != AF_INET)
817 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
819 t = IA_SIN(ia)->sin_addr;
825 * If the packet was transiting through us, use the address of
826 * the interface the packet came through in. If that interface
827 * doesn't have a suitable IP address, the normal selection
830 if (V_icmp_rfi && ifp != NULL) {
831 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
832 if (ifa->ifa_addr->sa_family != AF_INET)
835 t = IA_SIN(ia)->sin_addr;
840 * If the incoming packet was not addressed directly to us, use
841 * designated interface for icmp replies specified by sysctl
842 * net.inet.icmp.reply_src (default not set). Otherwise continue
843 * with normal source selection.
845 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
846 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
847 if (ifa->ifa_addr->sa_family != AF_INET)
850 t = IA_SIN(ia)->sin_addr;
855 * If the packet was transiting through us, use the address of
856 * the interface that is the closest to the packet source.
857 * When we don't have a route back to the packet source, stop here
858 * and drop the packet.
860 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
863 ICMPSTAT_INC(icps_noroute);
866 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr;
869 mac_netinet_icmp_replyinplace(m);
872 ip->ip_ttl = V_ip_defttl;
880 * Retrieve any source routing from the incoming packet;
881 * add on any record-route or timestamp options.
883 cp = (u_char *) (ip + 1);
884 if ((opts = ip_srcroute(m)) == NULL &&
885 (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
886 opts->m_len = sizeof(struct in_addr);
887 mtod(opts, struct in_addr *)->s_addr = 0;
892 printf("icmp_reflect optlen %d rt %d => ",
893 optlen, opts->m_len);
895 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
896 opt = cp[IPOPT_OPTVAL];
897 if (opt == IPOPT_EOL)
899 if (opt == IPOPT_NOP)
902 if (cnt < IPOPT_OLEN + sizeof(*cp))
904 len = cp[IPOPT_OLEN];
905 if (len < IPOPT_OLEN + sizeof(*cp) ||
910 * Should check for overflow, but it "can't happen"
912 if (opt == IPOPT_RR || opt == IPOPT_TS ||
913 opt == IPOPT_SECURITY) {
915 mtod(opts, caddr_t) + opts->m_len, len);
919 /* Terminate & pad, if necessary */
920 cnt = opts->m_len % 4;
922 for (; cnt < 4; cnt++) {
923 *(mtod(opts, caddr_t) + opts->m_len) =
930 printf("%d\n", opts->m_len);
935 m_tag_delete_nonpersistent(m);
936 m->m_flags &= ~(M_BCAST|M_MCAST);
944 * Verifies if redirect message is valid, according to RFC 1122
946 * @src: sockaddr with address of redirect originator
947 * @dst: sockaddr with destination in question
948 * @gateway: new proposed gateway
950 * Returns 0 on success.
953 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst,
954 struct sockaddr_in *gateway, u_int fibnum)
956 struct nhop_object *nh;
961 /* Verify the gateway is directly reachable. */
962 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL)
963 return (ENETUNREACH);
965 /* TODO: fib-aware. */
966 if (ifa_ifwithaddr_check((struct sockaddr *)gateway))
967 return (EHOSTUNREACH);
969 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0);
974 * If the redirect isn't from our current router for this dst,
975 * it's either old or wrong. If it redirects us to ourselves,
976 * we have a routing loop, perhaps as a result of an interface
977 * going down recently.
979 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa))
981 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK)
984 /* If host route already exists, ignore redirect. */
985 if (nh->nh_flags & NHF_HOST)
988 /* If the prefix is directly reachable, ignore redirect. */
989 if (!(nh->nh_flags & NHF_GATEWAY))
996 * Send an icmp packet back to the ip level,
997 * after supplying a checksum.
1000 icmp_send(struct mbuf *m, struct mbuf *opts)
1002 struct ip *ip = mtod(m, struct ip *);
1006 hlen = ip->ip_hl << 2;
1009 icp = mtod(m, struct icmp *);
1010 icp->icmp_cksum = 0;
1011 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
1014 m->m_pkthdr.rcvif = (struct ifnet *)0;
1017 char dstbuf[INET_ADDRSTRLEN];
1018 char srcbuf[INET_ADDRSTRLEN];
1020 printf("icmp_send dst %s src %s\n",
1021 inet_ntoa_r(ip->ip_dst, dstbuf),
1022 inet_ntoa_r(ip->ip_src, srcbuf));
1025 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
1029 * Return milliseconds since 00:00 UTC in network format.
1038 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1043 * Return the next larger or smaller MTU plateau (table from RFC 1191)
1044 * given current value MTU. If DIR is less than zero, a larger plateau
1045 * is returned; otherwise, a smaller value is returned.
1048 ip_next_mtu(int mtu, int dir)
1050 static int mtutab[] = {
1051 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
1056 size = (sizeof mtutab) / (sizeof mtutab[0]);
1058 for (i = 0; i < size; i++)
1059 if (mtu > mtutab[i])
1062 for (i = size - 1; i >= 0; i--)
1063 if (mtu < mtutab[i])
1065 if (mtu == mtutab[0])
1073 * badport_bandlim() - check for ICMP bandwidth limit
1075 * Return 0 if it is ok to send an ICMP error response, -1 if we have
1076 * hit our bandwidth limit and it is not ok.
1078 * If icmplim is <= 0, the feature is disabled and 0 is returned.
1080 * For now we separate the TCP and UDP subsystems w/ different 'which'
1081 * values. We may eventually remove this separation (and simplify the
1084 * Note that the printing of the error message is delayed so we can
1085 * properly print the icmp error rate that the system was trying to do
1086 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
1087 * the 'final' error, but it doesn't make sense to solve the printing
1088 * delay with more complex code.
1092 struct counter_rate cr;
1094 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
1095 { "icmp unreach response" },
1096 { "icmp ping response" },
1097 { "icmp tstamp response" },
1098 { "closed port RST response" },
1099 { "open port RST response" },
1100 { "icmp6 unreach response" },
1101 { "sctp ootb response" }
1103 #define V_icmp_rates VNET(icmp_rates)
1106 icmp_bandlimit_init(void)
1109 for (int i = 0; i < BANDLIM_MAX; i++) {
1110 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1111 V_icmp_rates[i].cr.cr_ticks = ticks;
1114 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1115 icmp_bandlimit_init, NULL);
1118 icmp_bandlimit_uninit(void)
1121 for (int i = 0; i < BANDLIM_MAX; i++)
1122 counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1124 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1125 icmp_bandlimit_uninit, NULL);
1128 badport_bandlim(int which)
1132 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1135 KASSERT(which >= 0 && which < BANDLIM_MAX,
1136 ("%s: which %d", __func__, which));
1138 if ((V_icmplim + V_icmplim_curr_jitter) <= 0)
1139 V_icmplim_curr_jitter = -V_icmplim + 1;
1141 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim +
1142 V_icmplim_curr_jitter);
1145 * Adjust limit +/- to jitter the measurement to deny a
1146 * side-channel port scan as in CVE-2020-25705
1148 if (V_icmplim_jitter > 0) {
1150 arc4random_uniform(V_icmplim_jitter * 2 +1)
1153 V_icmplim_curr_jitter = inc;
1158 if (pps > 0 && V_icmplim_output)
1159 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1160 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim +
1161 V_icmplim_curr_jitter);