2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
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>
55 #include <netinet/in.h>
56 #include <netinet/in_fib.h>
57 #include <netinet/in_pcb.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_icmp.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_options.h>
64 #include <netinet/sctp.h>
65 #include <netinet/tcp.h>
66 #include <netinet/tcp_var.h>
67 #include <netinet/tcpip.h>
68 #include <netinet/icmp_var.h>
72 #include <machine/in_cksum.h>
74 #include <security/mac/mac_framework.h>
78 * ICMP routines: error generation, receive packet processing, and
79 * routines to turnaround packets back to the originator, and
80 * host table maintenance routines.
82 static VNET_DEFINE(int, icmplim) = 200;
83 #define V_icmplim VNET(icmplim)
84 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
85 &VNET_NAME(icmplim), 0,
86 "Maximum number of ICMP responses per second");
88 static VNET_DEFINE(int, icmplim_output) = 1;
89 #define V_icmplim_output VNET(icmplim_output)
90 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
91 &VNET_NAME(icmplim_output), 0,
92 "Enable logging of ICMP response rate limiting");
95 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
96 VNET_PCPUSTAT_SYSINIT(icmpstat);
97 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
98 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
101 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
104 static VNET_DEFINE(int, icmpmaskrepl) = 0;
105 #define V_icmpmaskrepl VNET(icmpmaskrepl)
106 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
107 &VNET_NAME(icmpmaskrepl), 0,
108 "Reply to ICMP Address Mask Request packets");
110 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
111 #define V_icmpmaskfake VNET(icmpmaskfake)
112 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
113 &VNET_NAME(icmpmaskfake), 0,
114 "Fake reply to ICMP Address Mask Request packets");
116 VNET_DEFINE(int, drop_redirect) = 0;
117 #define V_drop_redirect VNET(drop_redirect)
118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
119 &VNET_NAME(drop_redirect), 0,
120 "Ignore ICMP redirects");
122 static VNET_DEFINE(int, log_redirect) = 0;
123 #define V_log_redirect VNET(log_redirect)
124 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
125 &VNET_NAME(log_redirect), 0,
126 "Log ICMP redirects to the console");
128 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
129 #define V_reply_src VNET(reply_src)
130 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
131 &VNET_NAME(reply_src), IFNAMSIZ,
132 "ICMP reply source for non-local packets");
134 static VNET_DEFINE(int, icmp_rfi) = 0;
135 #define V_icmp_rfi VNET(icmp_rfi)
136 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
137 &VNET_NAME(icmp_rfi), 0,
138 "ICMP reply from incoming interface for non-local packets");
140 static VNET_DEFINE(int, icmp_quotelen) = 8;
141 #define V_icmp_quotelen VNET(icmp_quotelen)
142 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
143 &VNET_NAME(icmp_quotelen), 0,
144 "Number of bytes from original packet to quote in ICMP reply");
146 static VNET_DEFINE(int, icmpbmcastecho) = 0;
147 #define V_icmpbmcastecho VNET(icmpbmcastecho)
148 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
149 &VNET_NAME(icmpbmcastecho), 0,
150 "Reply to multicast ICMP Echo Request and Timestamp packets");
152 static VNET_DEFINE(int, icmptstamprepl) = 1;
153 #define V_icmptstamprepl VNET(icmptstamprepl)
154 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
155 &VNET_NAME(icmptstamprepl), 0,
156 "Respond to ICMP Timestamp packets");
158 static VNET_DEFINE(int, error_keeptags) = 0;
159 #define V_error_keeptags VNET(error_keeptags)
160 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
161 &VNET_NAME(error_keeptags), 0,
162 "ICMP error response keeps copy of mbuf_tags of original packet");
168 static void icmp_reflect(struct mbuf *);
169 static void icmp_send(struct mbuf *, struct mbuf *);
171 extern struct protosw inetsw[];
174 * Kernel module interface for updating icmpstat. The argument is an index
175 * into icmpstat treated as an array of u_long. While this encodes the
176 * general layout of icmpstat into the caller, it doesn't encode its
177 * location, so that future changes to add, for example, per-CPU stats
178 * support won't cause binary compatibility problems for kernel modules.
181 kmod_icmpstat_inc(int statnum)
184 counter_u64_add(VNET(icmpstat)[statnum], 1);
188 * Generate an error packet of type error
189 * in response to bad packet ip.
192 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
194 struct ip *oip, *nip;
197 unsigned icmplen, icmpelen, nlen, oiphlen;
199 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
202 if (type != ICMP_REDIRECT)
203 ICMPSTAT_INC(icps_error);
206 * if the original packet was encrypted.
207 * if not the first fragment of message.
208 * in response to a multicast or broadcast packet.
209 * if the old packet protocol was an ICMP error message.
211 if (n->m_flags & M_DECRYPTED)
213 if (n->m_flags & (M_BCAST|M_MCAST))
216 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
217 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
219 oip = mtod(n, struct ip *);
220 oiphlen = oip->ip_hl << 2;
221 if (n->m_len < oiphlen + ICMP_MINLEN)
225 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
227 if (oip->ip_off & htons(~(IP_MF|IP_DF)))
229 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
230 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
231 oiphlen))->icmp_type)) {
232 ICMPSTAT_INC(icps_oldicmp);
236 * Calculate length to quote from original packet and
237 * prevent the ICMP mbuf from overflowing.
238 * Unfortunately this is non-trivial since ip_forward()
239 * sends us truncated packets.
241 nlen = m_length(n, NULL);
242 if (oip->ip_p == IPPROTO_TCP) {
246 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
249 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
250 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
252 oip = mtod(n, struct ip *);
253 th = mtodo(n, oiphlen);
254 tcphlen = th->th_off << 2;
255 if (tcphlen < sizeof(struct tcphdr))
257 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
259 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
261 if (n->m_len < oiphlen + tcphlen &&
262 (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
264 icmpelen = max(tcphlen, min(V_icmp_quotelen,
265 ntohs(oip->ip_len) - oiphlen));
266 } else if (oip->ip_p == IPPROTO_SCTP) {
268 struct sctp_chunkhdr *ch;
270 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
272 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
275 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
276 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
278 oip = mtod(n, struct ip *);
279 icmpelen = max(sizeof(struct sctphdr),
280 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
281 sh = mtodo(n, oiphlen);
282 if (ntohl(sh->v_tag) == 0 &&
283 ntohs(oip->ip_len) >= oiphlen +
284 sizeof(struct sctphdr) + 8 &&
285 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
286 n->m_next != NULL)) {
287 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
288 (n = m_pullup(n, oiphlen +
289 sizeof(struct sctphdr) + 8)) == NULL)
291 oip = mtod(n, struct ip *);
292 sh = mtodo(n, oiphlen);
293 ch = (struct sctp_chunkhdr *)(sh + 1);
294 if (ch->chunk_type == SCTP_INITIATION) {
295 icmpelen = max(sizeof(struct sctphdr) + 8,
296 min(V_icmp_quotelen, ntohs(oip->ip_len) -
301 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
304 icmplen = min(oiphlen + icmpelen, nlen);
305 if (icmplen < sizeof(struct ip))
308 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
309 m = m_gethdr(M_NOWAIT, MT_DATA);
311 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
315 mac_netinet_icmp_reply(n, m);
317 icmplen = min(icmplen, M_TRAILINGSPACE(m) -
318 sizeof(struct ip) - ICMP_MINLEN);
319 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
320 m->m_data += sizeof(struct ip);
321 m->m_len = ICMP_MINLEN + icmplen;
323 /* XXX MRT make the outgoing packet use the same FIB
324 * that was associated with the incoming packet
326 M_SETFIB(m, M_GETFIB(n));
327 icp = mtod(m, struct icmp *);
328 ICMPSTAT_INC(icps_outhist[type]);
329 icp->icmp_type = type;
330 if (type == ICMP_REDIRECT)
331 icp->icmp_gwaddr.s_addr = dest;
335 * The following assignments assume an overlay with the
336 * just zeroed icmp_void field.
338 if (type == ICMP_PARAMPROB) {
339 icp->icmp_pptr = code;
341 } else if (type == ICMP_UNREACH &&
342 code == ICMP_UNREACH_NEEDFRAG && mtu) {
343 icp->icmp_nextmtu = htons(mtu);
346 icp->icmp_code = code;
349 * Copy the quotation into ICMP message and
350 * convert quoted IP header back to network representation.
352 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
356 * Set up ICMP message mbuf and copy old IP header (without options
357 * in front of ICMP message.
358 * If the original mbuf was meant to bypass the firewall, the error
359 * reply should bypass as well.
361 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
362 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
363 ("insufficient space for ip header"));
364 m->m_data -= sizeof(struct ip);
365 m->m_len += sizeof(struct ip);
366 m->m_pkthdr.len = m->m_len;
367 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
368 nip = mtod(m, struct ip *);
369 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
370 nip->ip_len = htons(m->m_len);
371 nip->ip_v = IPVERSION;
373 nip->ip_p = IPPROTO_ICMP;
377 if (V_error_keeptags)
378 m_tag_copy_chain(m, n, M_NOWAIT);
387 * Process a received ICMP message.
390 icmp_input(struct mbuf **mp, int *offp, int proto)
393 struct in_ifaddr *ia;
394 struct mbuf *m = *mp;
395 struct ip *ip = mtod(m, struct ip *);
396 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
398 int icmplen = ntohs(ip->ip_len) - *offp;
400 void (*ctlfunc)(int, struct sockaddr *, void *);
406 * Locate icmp structure in mbuf, and check
407 * that not corrupted and of at least minimum length.
411 char srcbuf[INET_ADDRSTRLEN];
412 char dstbuf[INET_ADDRSTRLEN];
414 printf("icmp_input from %s to %s, len %d\n",
415 inet_ntoa_r(ip->ip_src, srcbuf),
416 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
419 if (icmplen < ICMP_MINLEN) {
420 ICMPSTAT_INC(icps_tooshort);
423 i = hlen + min(icmplen, ICMP_ADVLENMIN);
424 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
425 ICMPSTAT_INC(icps_tooshort);
426 return (IPPROTO_DONE);
428 ip = mtod(m, struct ip *);
431 icp = mtod(m, struct icmp *);
432 if (in_cksum(m, icmplen)) {
433 ICMPSTAT_INC(icps_checksum);
441 printf("icmp_input, type %d code %d\n", icp->icmp_type,
446 * Message type specific processing.
448 if (icp->icmp_type > ICMP_MAXTYPE)
452 bzero(&icmpsrc, sizeof(icmpsrc));
453 icmpsrc.sin_len = sizeof(struct sockaddr_in);
454 icmpsrc.sin_family = AF_INET;
455 bzero(&icmpdst, sizeof(icmpdst));
456 icmpdst.sin_len = sizeof(struct sockaddr_in);
457 icmpdst.sin_family = AF_INET;
458 bzero(&icmpgw, sizeof(icmpgw));
459 icmpgw.sin_len = sizeof(struct sockaddr_in);
460 icmpgw.sin_family = AF_INET;
462 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
463 code = icp->icmp_code;
464 switch (icp->icmp_type) {
468 case ICMP_UNREACH_NET:
469 case ICMP_UNREACH_HOST:
470 case ICMP_UNREACH_SRCFAIL:
471 case ICMP_UNREACH_NET_UNKNOWN:
472 case ICMP_UNREACH_HOST_UNKNOWN:
473 case ICMP_UNREACH_ISOLATED:
474 case ICMP_UNREACH_TOSNET:
475 case ICMP_UNREACH_TOSHOST:
476 case ICMP_UNREACH_HOST_PRECEDENCE:
477 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
478 code = PRC_UNREACH_NET;
481 case ICMP_UNREACH_NEEDFRAG:
486 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
487 * Treat subcodes 2,3 as immediate RST
489 case ICMP_UNREACH_PROTOCOL:
490 code = PRC_UNREACH_PROTOCOL;
492 case ICMP_UNREACH_PORT:
493 code = PRC_UNREACH_PORT;
496 case ICMP_UNREACH_NET_PROHIB:
497 case ICMP_UNREACH_HOST_PROHIB:
498 case ICMP_UNREACH_FILTER_PROHIB:
499 code = PRC_UNREACH_ADMIN_PROHIB;
510 code += PRC_TIMXCEED_INTRANS;
516 code = PRC_PARAMPROB;
519 * Problem with datagram; advise higher level routines.
521 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
522 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
523 ICMPSTAT_INC(icps_badlen);
526 /* Discard ICMP's in response to multicast packets */
527 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
531 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
533 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
535 * XXX if the packet contains [IPv4 AH TCP], we can't make a
536 * notification to TCP layer.
538 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
540 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
541 /* This should actually not happen */
542 ICMPSTAT_INC(icps_tooshort);
543 return (IPPROTO_DONE);
545 ip = mtod(m, struct ip *);
546 icp = (struct icmp *)(ip + 1);
548 * The upper layer handler can rely on:
549 * - The outer IP header has no options.
550 * - The outer IP header, the ICMP header, the inner IP header,
551 * and the first n bytes of the inner payload are contiguous.
552 * n is at least 8, but might be larger based on
553 * ICMP_ADVLENPREF. See its definition in ip_icmp.h.
555 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
557 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
558 (void *)&icp->icmp_ip);
562 ICMPSTAT_INC(icps_badcode);
566 if (!V_icmpbmcastecho
567 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
568 ICMPSTAT_INC(icps_bmcastecho);
571 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
573 icp->icmp_type = ICMP_ECHOREPLY;
577 if (V_icmptstamprepl == 0)
579 if (!V_icmpbmcastecho
580 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
581 ICMPSTAT_INC(icps_bmcasttstamp);
584 if (icmplen < ICMP_TSLEN) {
585 ICMPSTAT_INC(icps_badlen);
588 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
590 icp->icmp_type = ICMP_TSTAMPREPLY;
591 icp->icmp_rtime = iptime();
592 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
596 if (V_icmpmaskrepl == 0)
599 * We are not able to respond with all ones broadcast
600 * unless we receive it over a point-to-point interface.
602 if (icmplen < ICMP_MASKLEN)
604 switch (ip->ip_dst.s_addr) {
606 case INADDR_BROADCAST:
608 icmpdst.sin_addr = ip->ip_src;
612 icmpdst.sin_addr = ip->ip_dst;
614 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
615 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
618 if (ia->ia_ifp == NULL) {
619 ifa_free(&ia->ia_ifa);
622 icp->icmp_type = ICMP_MASKREPLY;
623 if (V_icmpmaskfake == 0)
624 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
626 icp->icmp_mask = V_icmpmaskfake;
627 if (ip->ip_src.s_addr == 0) {
628 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
629 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
630 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
631 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
633 ifa_free(&ia->ia_ifa);
635 ICMPSTAT_INC(icps_reflect);
636 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
638 return (IPPROTO_DONE);
641 if (V_log_redirect) {
644 src = ntohl(ip->ip_src.s_addr);
645 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
646 gw = ntohl(icp->icmp_gwaddr.s_addr);
647 printf("icmp redirect from %d.%d.%d.%d: "
648 "%d.%d.%d.%d => %d.%d.%d.%d\n",
649 (int)(src >> 24), (int)((src >> 16) & 0xff),
650 (int)((src >> 8) & 0xff), (int)(src & 0xff),
651 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
652 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
653 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
654 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
657 * RFC1812 says we must ignore ICMP redirects if we
658 * are acting as router.
660 if (V_drop_redirect || V_ipforwarding)
664 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
665 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
666 ICMPSTAT_INC(icps_badlen);
670 * Short circuit routing redirects to force
671 * immediate change in the kernel's routing
672 * tables. The message is also handed to anyone
673 * listening on a raw socket (e.g. the routing
674 * daemon for use in updating its tables).
676 icmpgw.sin_addr = ip->ip_src;
677 icmpdst.sin_addr = icp->icmp_gwaddr;
680 char dstbuf[INET_ADDRSTRLEN];
681 char gwbuf[INET_ADDRSTRLEN];
683 printf("redirect dst %s to %s\n",
684 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
685 inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
688 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
689 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
690 in_rtredirect((struct sockaddr *)&icmpsrc,
691 (struct sockaddr *)&icmpdst,
692 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
693 (struct sockaddr *)&icmpgw, fibnum);
695 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
699 * No kernel processing for the following;
700 * just fall through to send to raw listener.
703 case ICMP_ROUTERADVERT:
704 case ICMP_ROUTERSOLICIT:
705 case ICMP_TSTAMPREPLY:
708 case ICMP_SOURCEQUENCH:
715 rip_input(mp, offp, proto);
716 return (IPPROTO_DONE);
720 return (IPPROTO_DONE);
724 * Reflect the ip packet back to the source
727 icmp_reflect(struct mbuf *m)
729 struct rm_priotracker in_ifa_tracker;
730 struct ip *ip = mtod(m, struct ip *);
733 struct in_ifaddr *ia;
735 struct nhop4_extended nh_ext;
736 struct mbuf *opts = NULL;
737 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
739 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
740 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
741 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
742 m_freem(m); /* Bad return address */
743 ICMPSTAT_INC(icps_badaddr);
744 goto done; /* Ip_output() will check for broadcast */
748 ip->ip_dst = ip->ip_src;
751 * Source selection for ICMP replies:
753 * If the incoming packet was addressed directly to one of our
754 * own addresses, use dst as the src for the reply.
756 IN_IFADDR_RLOCK(&in_ifa_tracker);
757 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
758 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
759 t = IA_SIN(ia)->sin_addr;
760 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
764 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
767 * If the incoming packet was addressed to one of our broadcast
768 * addresses, use the first non-broadcast address which corresponds
769 * to the incoming interface.
771 ifp = m->m_pkthdr.rcvif;
772 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
774 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
775 if (ifa->ifa_addr->sa_family != AF_INET)
778 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
780 t = IA_SIN(ia)->sin_addr;
781 IF_ADDR_RUNLOCK(ifp);
785 IF_ADDR_RUNLOCK(ifp);
788 * If the packet was transiting through us, use the address of
789 * the interface the packet came through in. If that interface
790 * doesn't have a suitable IP address, the normal selection
793 if (V_icmp_rfi && ifp != NULL) {
795 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
796 if (ifa->ifa_addr->sa_family != AF_INET)
799 t = IA_SIN(ia)->sin_addr;
800 IF_ADDR_RUNLOCK(ifp);
803 IF_ADDR_RUNLOCK(ifp);
806 * If the incoming packet was not addressed directly to us, use
807 * designated interface for icmp replies specified by sysctl
808 * net.inet.icmp.reply_src (default not set). Otherwise continue
809 * with normal source selection.
811 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
813 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
814 if (ifa->ifa_addr->sa_family != AF_INET)
817 t = IA_SIN(ia)->sin_addr;
818 IF_ADDR_RUNLOCK(ifp);
821 IF_ADDR_RUNLOCK(ifp);
824 * If the packet was transiting through us, use the address of
825 * the interface that is the closest to the packet source.
826 * When we don't have a route back to the packet source, stop here
827 * and drop the packet.
829 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
831 ICMPSTAT_INC(icps_noroute);
837 mac_netinet_icmp_replyinplace(m);
840 ip->ip_ttl = V_ip_defttl;
848 * Retrieve any source routing from the incoming packet;
849 * add on any record-route or timestamp options.
851 cp = (u_char *) (ip + 1);
852 if ((opts = ip_srcroute(m)) == NULL &&
853 (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
854 opts->m_len = sizeof(struct in_addr);
855 mtod(opts, struct in_addr *)->s_addr = 0;
860 printf("icmp_reflect optlen %d rt %d => ",
861 optlen, opts->m_len);
863 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
864 opt = cp[IPOPT_OPTVAL];
865 if (opt == IPOPT_EOL)
867 if (opt == IPOPT_NOP)
870 if (cnt < IPOPT_OLEN + sizeof(*cp))
872 len = cp[IPOPT_OLEN];
873 if (len < IPOPT_OLEN + sizeof(*cp) ||
878 * Should check for overflow, but it "can't happen"
880 if (opt == IPOPT_RR || opt == IPOPT_TS ||
881 opt == IPOPT_SECURITY) {
883 mtod(opts, caddr_t) + opts->m_len, len);
887 /* Terminate & pad, if necessary */
888 cnt = opts->m_len % 4;
890 for (; cnt < 4; cnt++) {
891 *(mtod(opts, caddr_t) + opts->m_len) =
898 printf("%d\n", opts->m_len);
903 m_tag_delete_nonpersistent(m);
904 m->m_flags &= ~(M_BCAST|M_MCAST);
912 * Send an icmp packet back to the ip level,
913 * after supplying a checksum.
916 icmp_send(struct mbuf *m, struct mbuf *opts)
918 struct ip *ip = mtod(m, struct ip *);
922 hlen = ip->ip_hl << 2;
925 icp = mtod(m, struct icmp *);
927 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
930 m->m_pkthdr.rcvif = (struct ifnet *)0;
933 char dstbuf[INET_ADDRSTRLEN];
934 char srcbuf[INET_ADDRSTRLEN];
936 printf("icmp_send dst %s src %s\n",
937 inet_ntoa_r(ip->ip_dst, dstbuf),
938 inet_ntoa_r(ip->ip_src, srcbuf));
941 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
945 * Return milliseconds since 00:00 UTC in network format.
954 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
959 * Return the next larger or smaller MTU plateau (table from RFC 1191)
960 * given current value MTU. If DIR is less than zero, a larger plateau
961 * is returned; otherwise, a smaller value is returned.
964 ip_next_mtu(int mtu, int dir)
966 static int mtutab[] = {
967 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
972 size = (sizeof mtutab) / (sizeof mtutab[0]);
974 for (i = 0; i < size; i++)
978 for (i = size - 1; i >= 0; i--)
981 if (mtu == mtutab[0])
990 * badport_bandlim() - check for ICMP bandwidth limit
992 * Return 0 if it is ok to send an ICMP error response, -1 if we have
993 * hit our bandwidth limit and it is not ok.
995 * If icmplim is <= 0, the feature is disabled and 0 is returned.
997 * For now we separate the TCP and UDP subsystems w/ different 'which'
998 * values. We may eventually remove this separation (and simplify the
1001 * Note that the printing of the error message is delayed so we can
1002 * properly print the icmp error rate that the system was trying to do
1003 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
1004 * the 'final' error, but it doesn't make sense to solve the printing
1005 * delay with more complex code.
1009 badport_bandlim(int which)
1012 #define N(a) (sizeof (a) / sizeof (a[0]))
1013 static struct rate {
1015 struct timeval lasttime;
1017 } rates[BANDLIM_MAX+1] = {
1018 { "icmp unreach response" },
1019 { "icmp ping response" },
1020 { "icmp tstamp response" },
1021 { "closed port RST response" },
1022 { "open port RST response" },
1023 { "icmp6 unreach response" },
1024 { "sctp ootb response" }
1028 * Return ok status if feature disabled or argument out of range.
1030 if (V_icmplim > 0 && (u_int) which < N(rates)) {
1031 struct rate *r = &rates[which];
1032 int opps = r->curpps;
1034 if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
1035 return -1; /* discard packet */
1037 * If we've dropped below the threshold after having
1038 * rate-limited traffic print the message. This preserves
1039 * the previous behaviour at the expense of added complexity.
1041 if (V_icmplim_output && opps > V_icmplim)
1042 log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1043 r->type, opps, V_icmplim);
1045 return 0; /* okay to send packet */