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>
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>
75 #include <machine/in_cksum.h>
77 #include <security/mac/mac_framework.h>
81 * ICMP routines: error generation, receive packet processing, and
82 * routines to turnaround packets back to the originator, and
83 * host table maintenance routines.
85 VNET_DEFINE_STATIC(int, icmplim) = 200;
86 #define V_icmplim VNET(icmplim)
87 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
88 &VNET_NAME(icmplim), 0,
89 "Maximum number of ICMP responses per second");
91 VNET_DEFINE_STATIC(int, icmplim_output) = 1;
92 #define V_icmplim_output VNET(icmplim_output)
93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
94 &VNET_NAME(icmplim_output), 0,
95 "Enable logging of ICMP response rate limiting");
98 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
99 VNET_PCPUSTAT_SYSINIT(icmpstat);
100 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
101 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
104 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
107 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
108 #define V_icmpmaskrepl VNET(icmpmaskrepl)
109 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
110 &VNET_NAME(icmpmaskrepl), 0,
111 "Reply to ICMP Address Mask Request packets");
113 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
114 #define V_icmpmaskfake VNET(icmpmaskfake)
115 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
116 &VNET_NAME(icmpmaskfake), 0,
117 "Fake reply to ICMP Address Mask Request packets");
119 VNET_DEFINE(int, drop_redirect) = 0;
120 #define V_drop_redirect VNET(drop_redirect)
121 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
122 &VNET_NAME(drop_redirect), 0,
123 "Ignore ICMP redirects");
125 VNET_DEFINE_STATIC(int, log_redirect) = 0;
126 #define V_log_redirect VNET(log_redirect)
127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
128 &VNET_NAME(log_redirect), 0,
129 "Log ICMP redirects to the console");
131 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
132 #define V_reply_src VNET(reply_src)
133 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
134 &VNET_NAME(reply_src), IFNAMSIZ,
135 "ICMP reply source for non-local packets");
137 VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
138 #define V_icmp_rfi VNET(icmp_rfi)
139 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
140 &VNET_NAME(icmp_rfi), 0,
141 "ICMP reply from incoming interface for non-local packets");
142 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
143 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
144 #define V_icmp_quotelen VNET(icmp_quotelen)
145 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
146 &VNET_NAME(icmp_quotelen), 0,
147 "Number of bytes from original packet to quote in ICMP reply");
149 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
150 #define V_icmpbmcastecho VNET(icmpbmcastecho)
151 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
152 &VNET_NAME(icmpbmcastecho), 0,
153 "Reply to multicast ICMP Echo Request and Timestamp packets");
155 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
156 #define V_icmptstamprepl VNET(icmptstamprepl)
157 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
158 &VNET_NAME(icmptstamprepl), 0,
159 "Respond to ICMP Timestamp packets");
161 VNET_DEFINE_STATIC(int, error_keeptags) = 0;
162 #define V_error_keeptags VNET(error_keeptags)
163 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
164 &VNET_NAME(error_keeptags), 0,
165 "ICMP error response keeps copy of mbuf_tags of original packet");
171 static void icmp_reflect(struct mbuf *);
172 static void icmp_send(struct mbuf *, struct mbuf *);
174 extern struct protosw inetsw[];
177 * Kernel module interface for updating icmpstat. The argument is an index
178 * into icmpstat treated as an array of u_long. While this encodes the
179 * general layout of icmpstat into the caller, it doesn't encode its
180 * location, so that future changes to add, for example, per-CPU stats
181 * support won't cause binary compatibility problems for kernel modules.
184 kmod_icmpstat_inc(int statnum)
187 counter_u64_add(VNET(icmpstat)[statnum], 1);
191 * Generate an error packet of type error
192 * in response to bad packet ip.
195 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
197 struct ip *oip, *nip;
200 unsigned icmplen, icmpelen, nlen, oiphlen;
202 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
205 if (type != ICMP_REDIRECT)
206 ICMPSTAT_INC(icps_error);
209 * if the original packet was encrypted.
210 * if not the first fragment of message.
211 * in response to a multicast or broadcast packet.
212 * if the old packet protocol was an ICMP error message.
214 if (n->m_flags & M_DECRYPTED)
216 if (n->m_flags & (M_BCAST|M_MCAST))
219 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
220 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
222 oip = mtod(n, struct ip *);
223 oiphlen = oip->ip_hl << 2;
224 if (n->m_len < oiphlen + ICMP_MINLEN)
228 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
230 if (oip->ip_off & htons(~(IP_MF|IP_DF)))
232 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
233 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
234 oiphlen))->icmp_type)) {
235 ICMPSTAT_INC(icps_oldicmp);
239 * Calculate length to quote from original packet and
240 * prevent the ICMP mbuf from overflowing.
241 * Unfortunately this is non-trivial since ip_forward()
242 * sends us truncated packets.
244 nlen = m_length(n, NULL);
245 if (oip->ip_p == IPPROTO_TCP) {
249 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
252 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
253 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
255 oip = mtod(n, struct ip *);
256 th = mtodo(n, oiphlen);
257 tcphlen = th->th_off << 2;
258 if (tcphlen < sizeof(struct tcphdr))
260 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
262 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
264 if (n->m_len < oiphlen + tcphlen &&
265 (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
267 oip = mtod(n, struct ip *);
268 icmpelen = max(tcphlen, min(V_icmp_quotelen,
269 ntohs(oip->ip_len) - oiphlen));
270 } else if (oip->ip_p == IPPROTO_SCTP) {
272 struct sctp_chunkhdr *ch;
274 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
276 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
279 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
280 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
282 oip = mtod(n, struct ip *);
283 icmpelen = max(sizeof(struct sctphdr),
284 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
285 sh = mtodo(n, oiphlen);
286 if (ntohl(sh->v_tag) == 0 &&
287 ntohs(oip->ip_len) >= oiphlen +
288 sizeof(struct sctphdr) + 8 &&
289 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
290 n->m_next != NULL)) {
291 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
292 (n = m_pullup(n, oiphlen +
293 sizeof(struct sctphdr) + 8)) == NULL)
295 oip = mtod(n, struct ip *);
296 sh = mtodo(n, oiphlen);
297 ch = (struct sctp_chunkhdr *)(sh + 1);
298 if (ch->chunk_type == SCTP_INITIATION) {
299 icmpelen = max(sizeof(struct sctphdr) + 8,
300 min(V_icmp_quotelen, ntohs(oip->ip_len) -
305 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
308 icmplen = min(oiphlen + icmpelen, nlen);
309 if (icmplen < sizeof(struct ip))
312 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
313 m = m_gethdr(M_NOWAIT, MT_DATA);
315 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
319 mac_netinet_icmp_reply(n, m);
321 icmplen = min(icmplen, M_TRAILINGSPACE(m) -
322 sizeof(struct ip) - ICMP_MINLEN);
323 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
324 m->m_data += sizeof(struct ip);
325 m->m_len = ICMP_MINLEN + icmplen;
327 /* XXX MRT make the outgoing packet use the same FIB
328 * that was associated with the incoming packet
330 M_SETFIB(m, M_GETFIB(n));
331 icp = mtod(m, struct icmp *);
332 ICMPSTAT_INC(icps_outhist[type]);
333 icp->icmp_type = type;
334 if (type == ICMP_REDIRECT)
335 icp->icmp_gwaddr.s_addr = dest;
339 * The following assignments assume an overlay with the
340 * just zeroed icmp_void field.
342 if (type == ICMP_PARAMPROB) {
343 icp->icmp_pptr = code;
345 } else if (type == ICMP_UNREACH &&
346 code == ICMP_UNREACH_NEEDFRAG && mtu) {
347 icp->icmp_nextmtu = htons(mtu);
350 icp->icmp_code = code;
353 * Copy the quotation into ICMP message and
354 * convert quoted IP header back to network representation.
356 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
360 * Set up ICMP message mbuf and copy old IP header (without options
361 * in front of ICMP message.
362 * If the original mbuf was meant to bypass the firewall, the error
363 * reply should bypass as well.
365 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
366 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
367 ("insufficient space for ip header"));
368 m->m_data -= sizeof(struct ip);
369 m->m_len += sizeof(struct ip);
370 m->m_pkthdr.len = m->m_len;
371 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
372 nip = mtod(m, struct ip *);
373 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
374 nip->ip_len = htons(m->m_len);
375 nip->ip_v = IPVERSION;
377 nip->ip_p = IPPROTO_ICMP;
381 if (V_error_keeptags)
382 m_tag_copy_chain(m, n, M_NOWAIT);
391 * Process a received ICMP message.
394 icmp_input(struct mbuf **mp, int *offp, int proto)
397 struct in_ifaddr *ia;
398 struct mbuf *m = *mp;
399 struct ip *ip = mtod(m, struct ip *);
400 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
402 int icmplen = ntohs(ip->ip_len) - *offp;
404 void (*ctlfunc)(int, struct sockaddr *, void *);
412 * Locate icmp structure in mbuf, and check
413 * that not corrupted and of at least minimum length.
417 char srcbuf[INET_ADDRSTRLEN];
418 char dstbuf[INET_ADDRSTRLEN];
420 printf("icmp_input from %s to %s, len %d\n",
421 inet_ntoa_r(ip->ip_src, srcbuf),
422 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
425 if (icmplen < ICMP_MINLEN) {
426 ICMPSTAT_INC(icps_tooshort);
429 i = hlen + min(icmplen, ICMP_ADVLENMIN);
430 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
431 ICMPSTAT_INC(icps_tooshort);
432 return (IPPROTO_DONE);
434 ip = mtod(m, struct ip *);
437 icp = mtod(m, struct icmp *);
438 if (in_cksum(m, icmplen)) {
439 ICMPSTAT_INC(icps_checksum);
447 printf("icmp_input, type %d code %d\n", icp->icmp_type,
452 * Message type specific processing.
454 if (icp->icmp_type > ICMP_MAXTYPE)
458 bzero(&icmpsrc, sizeof(icmpsrc));
459 icmpsrc.sin_len = sizeof(struct sockaddr_in);
460 icmpsrc.sin_family = AF_INET;
461 bzero(&icmpdst, sizeof(icmpdst));
462 icmpdst.sin_len = sizeof(struct sockaddr_in);
463 icmpdst.sin_family = AF_INET;
464 bzero(&icmpgw, sizeof(icmpgw));
465 icmpgw.sin_len = sizeof(struct sockaddr_in);
466 icmpgw.sin_family = AF_INET;
468 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
469 code = icp->icmp_code;
470 switch (icp->icmp_type) {
474 case ICMP_UNREACH_NET:
475 case ICMP_UNREACH_HOST:
476 case ICMP_UNREACH_SRCFAIL:
477 case ICMP_UNREACH_NET_UNKNOWN:
478 case ICMP_UNREACH_HOST_UNKNOWN:
479 case ICMP_UNREACH_ISOLATED:
480 case ICMP_UNREACH_TOSNET:
481 case ICMP_UNREACH_TOSHOST:
482 case ICMP_UNREACH_HOST_PRECEDENCE:
483 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
484 code = PRC_UNREACH_NET;
487 case ICMP_UNREACH_NEEDFRAG:
492 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
493 * Treat subcodes 2,3 as immediate RST
495 case ICMP_UNREACH_PROTOCOL:
496 code = PRC_UNREACH_PROTOCOL;
498 case ICMP_UNREACH_PORT:
499 code = PRC_UNREACH_PORT;
502 case ICMP_UNREACH_NET_PROHIB:
503 case ICMP_UNREACH_HOST_PROHIB:
504 case ICMP_UNREACH_FILTER_PROHIB:
505 code = PRC_UNREACH_ADMIN_PROHIB;
516 code += PRC_TIMXCEED_INTRANS;
522 code = PRC_PARAMPROB;
525 * Problem with datagram; advise higher level routines.
527 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
528 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
529 ICMPSTAT_INC(icps_badlen);
532 /* Discard ICMP's in response to multicast packets */
533 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
537 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
539 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
541 * XXX if the packet contains [IPv4 AH TCP], we can't make a
542 * notification to TCP layer.
544 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
546 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
547 /* This should actually not happen */
548 ICMPSTAT_INC(icps_tooshort);
549 return (IPPROTO_DONE);
551 ip = mtod(m, struct ip *);
552 icp = (struct icmp *)(ip + 1);
554 * The upper layer handler can rely on:
555 * - The outer IP header has no options.
556 * - The outer IP header, the ICMP header, the inner IP header,
557 * and the first n bytes of the inner payload are contiguous.
558 * n is at least 8, but might be larger based on
559 * ICMP_ADVLENPREF. See its definition in ip_icmp.h.
561 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
563 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
564 (void *)&icp->icmp_ip);
568 ICMPSTAT_INC(icps_badcode);
572 if (!V_icmpbmcastecho
573 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
574 ICMPSTAT_INC(icps_bmcastecho);
577 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
579 icp->icmp_type = ICMP_ECHOREPLY;
583 if (V_icmptstamprepl == 0)
585 if (!V_icmpbmcastecho
586 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
587 ICMPSTAT_INC(icps_bmcasttstamp);
590 if (icmplen < ICMP_TSLEN) {
591 ICMPSTAT_INC(icps_badlen);
594 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
596 icp->icmp_type = ICMP_TSTAMPREPLY;
597 icp->icmp_rtime = iptime();
598 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
602 if (V_icmpmaskrepl == 0)
605 * We are not able to respond with all ones broadcast
606 * unless we receive it over a point-to-point interface.
608 if (icmplen < ICMP_MASKLEN)
610 switch (ip->ip_dst.s_addr) {
612 case INADDR_BROADCAST:
614 icmpdst.sin_addr = ip->ip_src;
618 icmpdst.sin_addr = ip->ip_dst;
620 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
621 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
624 if (ia->ia_ifp == NULL)
626 icp->icmp_type = ICMP_MASKREPLY;
627 if (V_icmpmaskfake == 0)
628 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
630 icp->icmp_mask = V_icmpmaskfake;
631 if (ip->ip_src.s_addr == 0) {
632 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
633 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
634 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
635 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
638 ICMPSTAT_INC(icps_reflect);
639 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
641 return (IPPROTO_DONE);
644 if (V_log_redirect) {
647 src = ntohl(ip->ip_src.s_addr);
648 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
649 gw = ntohl(icp->icmp_gwaddr.s_addr);
650 printf("icmp redirect from %d.%d.%d.%d: "
651 "%d.%d.%d.%d => %d.%d.%d.%d\n",
652 (int)(src >> 24), (int)((src >> 16) & 0xff),
653 (int)((src >> 8) & 0xff), (int)(src & 0xff),
654 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
655 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
656 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
657 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
660 * RFC1812 says we must ignore ICMP redirects if we
661 * are acting as router.
663 if (V_drop_redirect || V_ipforwarding)
667 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
668 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
669 ICMPSTAT_INC(icps_badlen);
673 * Short circuit routing redirects to force
674 * immediate change in the kernel's routing
675 * tables. The message is also handed to anyone
676 * listening on a raw socket (e.g. the routing
677 * daemon for use in updating its tables).
679 icmpgw.sin_addr = ip->ip_src;
680 icmpdst.sin_addr = icp->icmp_gwaddr;
683 char dstbuf[INET_ADDRSTRLEN];
684 char gwbuf[INET_ADDRSTRLEN];
686 printf("redirect dst %s to %s\n",
687 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
688 inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
691 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
692 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
693 in_rtredirect((struct sockaddr *)&icmpsrc,
694 (struct sockaddr *)&icmpdst,
695 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
696 (struct sockaddr *)&icmpgw, fibnum);
698 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
702 * No kernel processing for the following;
703 * just fall through to send to raw listener.
706 case ICMP_ROUTERADVERT:
707 case ICMP_ROUTERSOLICIT:
708 case ICMP_TSTAMPREPLY:
711 case ICMP_SOURCEQUENCH:
718 rip_input(mp, offp, proto);
719 return (IPPROTO_DONE);
723 return (IPPROTO_DONE);
727 * Reflect the ip packet back to the source
730 icmp_reflect(struct mbuf *m)
732 struct rm_priotracker in_ifa_tracker;
733 struct ip *ip = mtod(m, struct ip *);
736 struct in_ifaddr *ia;
738 struct nhop4_extended nh_ext;
739 struct mbuf *opts = NULL;
740 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
744 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
745 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
746 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
747 m_freem(m); /* Bad return address */
748 ICMPSTAT_INC(icps_badaddr);
749 goto done; /* Ip_output() will check for broadcast */
753 ip->ip_dst = ip->ip_src;
756 * Source selection for ICMP replies:
758 * If the incoming packet was addressed directly to one of our
759 * own addresses, use dst as the src for the reply.
761 IN_IFADDR_RLOCK(&in_ifa_tracker);
762 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
763 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
764 t = IA_SIN(ia)->sin_addr;
765 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
769 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
772 * If the incoming packet was addressed to one of our broadcast
773 * addresses, use the first non-broadcast address which corresponds
774 * to the incoming interface.
776 ifp = m->m_pkthdr.rcvif;
777 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
778 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
779 if (ifa->ifa_addr->sa_family != AF_INET)
782 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
784 t = IA_SIN(ia)->sin_addr;
790 * If the packet was transiting through us, use the address of
791 * the interface the packet came through in. If that interface
792 * doesn't have a suitable IP address, the normal selection
795 if (V_icmp_rfi && ifp != NULL) {
796 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
797 if (ifa->ifa_addr->sa_family != AF_INET)
800 t = IA_SIN(ia)->sin_addr;
805 * If the incoming packet was not addressed directly to us, use
806 * designated interface for icmp replies specified by sysctl
807 * net.inet.icmp.reply_src (default not set). Otherwise continue
808 * with normal source selection.
810 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
811 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
812 if (ifa->ifa_addr->sa_family != AF_INET)
815 t = IA_SIN(ia)->sin_addr;
820 * If the packet was transiting through us, use the address of
821 * the interface that is the closest to the packet source.
822 * When we don't have a route back to the packet source, stop here
823 * and drop the packet.
825 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
827 ICMPSTAT_INC(icps_noroute);
833 mac_netinet_icmp_replyinplace(m);
836 ip->ip_ttl = V_ip_defttl;
844 * Retrieve any source routing from the incoming packet;
845 * add on any record-route or timestamp options.
847 cp = (u_char *) (ip + 1);
848 if ((opts = ip_srcroute(m)) == NULL &&
849 (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
850 opts->m_len = sizeof(struct in_addr);
851 mtod(opts, struct in_addr *)->s_addr = 0;
856 printf("icmp_reflect optlen %d rt %d => ",
857 optlen, opts->m_len);
859 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
860 opt = cp[IPOPT_OPTVAL];
861 if (opt == IPOPT_EOL)
863 if (opt == IPOPT_NOP)
866 if (cnt < IPOPT_OLEN + sizeof(*cp))
868 len = cp[IPOPT_OLEN];
869 if (len < IPOPT_OLEN + sizeof(*cp) ||
874 * Should check for overflow, but it "can't happen"
876 if (opt == IPOPT_RR || opt == IPOPT_TS ||
877 opt == IPOPT_SECURITY) {
879 mtod(opts, caddr_t) + opts->m_len, len);
883 /* Terminate & pad, if necessary */
884 cnt = opts->m_len % 4;
886 for (; cnt < 4; cnt++) {
887 *(mtod(opts, caddr_t) + opts->m_len) =
894 printf("%d\n", opts->m_len);
899 m_tag_delete_nonpersistent(m);
900 m->m_flags &= ~(M_BCAST|M_MCAST);
908 * Send an icmp packet back to the ip level,
909 * after supplying a checksum.
912 icmp_send(struct mbuf *m, struct mbuf *opts)
914 struct ip *ip = mtod(m, struct ip *);
918 hlen = ip->ip_hl << 2;
921 icp = mtod(m, struct icmp *);
923 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
926 m->m_pkthdr.rcvif = (struct ifnet *)0;
929 char dstbuf[INET_ADDRSTRLEN];
930 char srcbuf[INET_ADDRSTRLEN];
932 printf("icmp_send dst %s src %s\n",
933 inet_ntoa_r(ip->ip_dst, dstbuf),
934 inet_ntoa_r(ip->ip_src, srcbuf));
937 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
941 * Return milliseconds since 00:00 UTC in network format.
950 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
955 * Return the next larger or smaller MTU plateau (table from RFC 1191)
956 * given current value MTU. If DIR is less than zero, a larger plateau
957 * is returned; otherwise, a smaller value is returned.
960 ip_next_mtu(int mtu, int dir)
962 static int mtutab[] = {
963 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
968 size = (sizeof mtutab) / (sizeof mtutab[0]);
970 for (i = 0; i < size; i++)
974 for (i = size - 1; i >= 0; i--)
977 if (mtu == mtutab[0])
986 * badport_bandlim() - check for ICMP bandwidth limit
988 * Return 0 if it is ok to send an ICMP error response, -1 if we have
989 * hit our bandwidth limit and it is not ok.
991 * If icmplim is <= 0, the feature is disabled and 0 is returned.
993 * For now we separate the TCP and UDP subsystems w/ different 'which'
994 * values. We may eventually remove this separation (and simplify the
997 * Note that the printing of the error message is delayed so we can
998 * properly print the icmp error rate that the system was trying to do
999 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
1000 * the 'final' error, but it doesn't make sense to solve the printing
1001 * delay with more complex code.
1005 struct counter_rate cr;
1007 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
1008 { "icmp unreach response" },
1009 { "icmp ping response" },
1010 { "icmp tstamp response" },
1011 { "closed port RST response" },
1012 { "open port RST response" },
1013 { "icmp6 unreach response" },
1014 { "sctp ootb response" }
1016 #define V_icmp_rates VNET(icmp_rates)
1019 icmp_bandlimit_init(void)
1022 for (int i = 0; i < BANDLIM_MAX; i++) {
1023 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1024 V_icmp_rates[i].cr.cr_ticks = ticks;
1027 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1028 icmp_bandlimit_init, NULL);
1031 icmp_bandlimit_uninit(void)
1034 for (int i = 0; i < BANDLIM_MAX; i++)
1035 counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1037 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1038 icmp_bandlimit_uninit, NULL);
1041 badport_bandlim(int which)
1045 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1048 KASSERT(which >= 0 && which < BANDLIM_MAX,
1049 ("%s: which %d", __func__, which));
1051 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim);
1054 if (pps > 0 && V_icmplim_output)
1055 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1056 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim);