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("$FreeBSD$");
35 #include "opt_ipsec.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>
44 #include <sys/sysctl.h>
45 #include <sys/syslog.h>
48 #include <net/if_types.h>
49 #include <net/route.h>
52 #include <netinet/in.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip.h>
57 #include <netinet/ip_icmp.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/ip_options.h>
60 #include <netinet/tcp.h>
61 #include <netinet/tcp_var.h>
62 #include <netinet/tcpip.h>
63 #include <netinet/icmp_var.h>
66 #include <netipsec/ipsec.h>
67 #include <netipsec/key.h>
70 #include <machine/in_cksum.h>
72 #include <security/mac/mac_framework.h>
75 * ICMP routines: error generation, receive packet processing, and
76 * routines to turnaround packets back to the originator, and
77 * host table maintenance routines.
79 VNET_DEFINE(struct icmpstat, icmpstat);
80 SYSCTL_VNET_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
81 &VNET_NAME(icmpstat), icmpstat, "");
83 static VNET_DEFINE(int, icmpmaskrepl) = 0;
84 #define V_icmpmaskrepl VNET(icmpmaskrepl)
85 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
86 &VNET_NAME(icmpmaskrepl), 0,
87 "Reply to ICMP Address Mask Request packets.");
89 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
90 #define V_icmpmaskfake VNET(icmpmaskfake)
91 SYSCTL_VNET_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
92 &VNET_NAME(icmpmaskfake), 0,
93 "Fake reply to ICMP Address Mask Request packets.");
95 VNET_DEFINE(int, drop_redirect) = 0;
97 static VNET_DEFINE(int, log_redirect) = 0;
98 #define V_log_redirect VNET(log_redirect)
99 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
100 &VNET_NAME(log_redirect), 0,
101 "Log ICMP redirects to the console");
103 static VNET_DEFINE(int, icmplim) = 200;
104 #define V_icmplim VNET(icmplim)
105 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
106 &VNET_NAME(icmplim), 0,
107 "Maximum number of ICMP responses per second");
109 static VNET_DEFINE(int, icmplim_output) = 1;
110 #define V_icmplim_output VNET(icmplim_output)
111 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
112 &VNET_NAME(icmplim_output), 0,
113 "Enable rate limiting of ICMP responses");
115 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
116 #define V_reply_src VNET(reply_src)
117 SYSCTL_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
118 &VNET_NAME(reply_src), IFNAMSIZ,
119 "icmp reply source for non-local packets.");
121 static VNET_DEFINE(int, icmp_rfi) = 0;
122 #define V_icmp_rfi VNET(icmp_rfi)
123 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
124 &VNET_NAME(icmp_rfi), 0,
125 "ICMP reply from incoming interface for non-local packets");
127 static VNET_DEFINE(int, icmp_quotelen) = 8;
128 #define V_icmp_quotelen VNET(icmp_quotelen)
129 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
130 &VNET_NAME(icmp_quotelen), 0,
131 "Number of bytes from original packet to quote in ICMP reply");
134 * ICMP broadcast echo sysctl
136 static VNET_DEFINE(int, icmpbmcastecho) = 0;
137 #define V_icmpbmcastecho VNET(icmpbmcastecho)
138 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
139 &VNET_NAME(icmpbmcastecho), 0,
147 static void icmp_reflect(struct mbuf *);
148 static void icmp_send(struct mbuf *, struct mbuf *);
150 extern struct protosw inetsw[];
153 sysctl_net_icmp_drop_redir(SYSCTL_HANDLER_ARGS)
157 struct radix_node_head *rnh;
159 new = V_drop_redirect;
160 error = sysctl_handle_int(oidp, &new, 0, req);
161 if (error == 0 && req->newptr) {
162 new = (new != 0) ? 1 : 0;
164 if (new == V_drop_redirect)
167 for (i = 0; i < rt_numfibs; i++) {
168 if ((rnh = rt_tables_get_rnh(i, AF_INET)) == NULL)
170 RADIX_NODE_HEAD_LOCK(rnh);
171 in_setmatchfunc(rnh, new);
172 RADIX_NODE_HEAD_UNLOCK(rnh);
175 V_drop_redirect = new;
181 SYSCTL_VNET_PROC(_net_inet_icmp, OID_AUTO, drop_redirect,
182 CTLTYPE_INT|CTLFLAG_RW, 0, 0,
183 sysctl_net_icmp_drop_redir, "I", "Ignore ICMP redirects");
186 * Kernel module interface for updating icmpstat. The argument is an index
187 * into icmpstat treated as an array of u_long. While this encodes the
188 * general layout of icmpstat into the caller, it doesn't encode its
189 * location, so that future changes to add, for example, per-CPU stats
190 * support won't cause binary compatibility problems for kernel modules.
193 kmod_icmpstat_inc(int statnum)
196 (*((u_long *)&V_icmpstat + statnum))++;
200 * Generate an error packet of type error
201 * in response to bad packet ip.
204 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
206 register struct ip *oip = mtod(n, struct ip *), *nip;
207 register unsigned oiphlen = oip->ip_hl << 2;
208 register struct icmp *icp;
209 register struct mbuf *m;
210 unsigned icmplen, icmpelen, nlen;
212 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
215 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
217 if (type != ICMP_REDIRECT)
218 ICMPSTAT_INC(icps_error);
221 * if the original packet was encrypted.
222 * if not the first fragment of message.
223 * in response to a multicast or broadcast packet.
224 * if the old packet protocol was an ICMP error message.
226 if (n->m_flags & M_DECRYPTED)
228 if (oip->ip_off & ~(IP_MF|IP_DF))
230 if (n->m_flags & (M_BCAST|M_MCAST))
232 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
233 n->m_len >= oiphlen + ICMP_MINLEN &&
234 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
235 ICMPSTAT_INC(icps_oldicmp);
238 /* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
239 if (oiphlen + 8 > n->m_len)
242 * Calculate length to quote from original packet and
243 * prevent the ICMP mbuf from overflowing.
244 * Unfortunatly this is non-trivial since ip_forward()
245 * sends us truncated packets.
247 nlen = m_length(n, NULL);
248 if (oip->ip_p == IPPROTO_TCP) {
252 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
255 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
256 ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
258 th = (struct tcphdr *)((caddr_t)oip + oiphlen);
259 tcphlen = th->th_off << 2;
260 if (tcphlen < sizeof(struct tcphdr))
262 if (oip->ip_len < oiphlen + tcphlen)
264 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
266 if (n->m_len < oiphlen + tcphlen &&
267 ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
269 icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
271 stdreply: icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
273 icmplen = min(oiphlen + icmpelen, nlen);
274 if (icmplen < sizeof(struct ip))
277 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
278 m = m_gethdr(M_DONTWAIT, MT_DATA);
280 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
284 mac_netinet_icmp_reply(n, m);
286 icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
287 m_align(m, ICMP_MINLEN + icmplen);
288 m->m_len = ICMP_MINLEN + icmplen;
290 /* XXX MRT make the outgoing packet use the same FIB
291 * that was associated with the incoming packet
293 M_SETFIB(m, M_GETFIB(n));
294 icp = mtod(m, struct icmp *);
295 ICMPSTAT_INC(icps_outhist[type]);
296 icp->icmp_type = type;
297 if (type == ICMP_REDIRECT)
298 icp->icmp_gwaddr.s_addr = dest;
302 * The following assignments assume an overlay with the
303 * just zeroed icmp_void field.
305 if (type == ICMP_PARAMPROB) {
306 icp->icmp_pptr = code;
308 } else if (type == ICMP_UNREACH &&
309 code == ICMP_UNREACH_NEEDFRAG && mtu) {
310 icp->icmp_nextmtu = htons(mtu);
313 icp->icmp_code = code;
316 * Copy the quotation into ICMP message and
317 * convert quoted IP header back to network representation.
319 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
321 nip->ip_len = htons(nip->ip_len);
322 nip->ip_off = htons(nip->ip_off);
325 * Set up ICMP message mbuf and copy old IP header (without options
326 * in front of ICMP message.
327 * If the original mbuf was meant to bypass the firewall, the error
328 * reply should bypass as well.
330 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
331 m->m_data -= sizeof(struct ip);
332 m->m_len += sizeof(struct ip);
333 m->m_pkthdr.len = m->m_len;
334 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
335 nip = mtod(m, struct ip *);
336 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
337 nip->ip_len = m->m_len;
338 nip->ip_v = IPVERSION;
340 nip->ip_p = IPPROTO_ICMP;
350 * Process a received ICMP message.
353 icmp_input(struct mbuf *m, int off)
356 struct in_ifaddr *ia;
357 struct ip *ip = mtod(m, struct ip *);
358 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
360 int icmplen = ip->ip_len;
362 void (*ctlfunc)(int, struct sockaddr *, void *);
366 * Locate icmp structure in mbuf, and check
367 * that not corrupted and of at least minimum length.
371 char buf[4 * sizeof "123"];
372 strcpy(buf, inet_ntoa(ip->ip_src));
373 printf("icmp_input from %s to %s, len %d\n",
374 buf, inet_ntoa(ip->ip_dst), icmplen);
377 if (icmplen < ICMP_MINLEN) {
378 ICMPSTAT_INC(icps_tooshort);
381 i = hlen + min(icmplen, ICMP_ADVLENMIN);
382 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
383 ICMPSTAT_INC(icps_tooshort);
386 ip = mtod(m, struct ip *);
389 icp = mtod(m, struct icmp *);
390 if (in_cksum(m, icmplen)) {
391 ICMPSTAT_INC(icps_checksum);
397 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
399 * Deliver very specific ICMP type only.
401 switch (icp->icmp_type) {
412 printf("icmp_input, type %d code %d\n", icp->icmp_type,
417 * Message type specific processing.
419 if (icp->icmp_type > ICMP_MAXTYPE)
423 bzero(&icmpsrc, sizeof(icmpsrc));
424 icmpsrc.sin_len = sizeof(struct sockaddr_in);
425 icmpsrc.sin_family = AF_INET;
426 bzero(&icmpdst, sizeof(icmpdst));
427 icmpdst.sin_len = sizeof(struct sockaddr_in);
428 icmpdst.sin_family = AF_INET;
429 bzero(&icmpgw, sizeof(icmpgw));
430 icmpgw.sin_len = sizeof(struct sockaddr_in);
431 icmpgw.sin_family = AF_INET;
433 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
434 code = icp->icmp_code;
435 switch (icp->icmp_type) {
439 case ICMP_UNREACH_NET:
440 case ICMP_UNREACH_HOST:
441 case ICMP_UNREACH_SRCFAIL:
442 case ICMP_UNREACH_NET_UNKNOWN:
443 case ICMP_UNREACH_HOST_UNKNOWN:
444 case ICMP_UNREACH_ISOLATED:
445 case ICMP_UNREACH_TOSNET:
446 case ICMP_UNREACH_TOSHOST:
447 case ICMP_UNREACH_HOST_PRECEDENCE:
448 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
449 code = PRC_UNREACH_NET;
452 case ICMP_UNREACH_NEEDFRAG:
457 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
458 * Treat subcodes 2,3 as immediate RST
460 case ICMP_UNREACH_PROTOCOL:
461 case ICMP_UNREACH_PORT:
462 code = PRC_UNREACH_PORT;
465 case ICMP_UNREACH_NET_PROHIB:
466 case ICMP_UNREACH_HOST_PROHIB:
467 case ICMP_UNREACH_FILTER_PROHIB:
468 code = PRC_UNREACH_ADMIN_PROHIB;
479 code += PRC_TIMXCEED_INTRANS;
485 code = PRC_PARAMPROB;
488 case ICMP_SOURCEQUENCH:
494 * Problem with datagram; advise higher level routines.
496 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
497 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
498 ICMPSTAT_INC(icps_badlen);
501 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
502 /* Discard ICMP's in response to multicast packets */
503 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
507 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
509 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
511 * XXX if the packet contains [IPv4 AH TCP], we can't make a
512 * notification to TCP layer.
514 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
516 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
517 (void *)&icp->icmp_ip);
521 ICMPSTAT_INC(icps_badcode);
525 if (!V_icmpbmcastecho
526 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
527 ICMPSTAT_INC(icps_bmcastecho);
530 icp->icmp_type = ICMP_ECHOREPLY;
531 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
537 if (!V_icmpbmcastecho
538 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
539 ICMPSTAT_INC(icps_bmcasttstamp);
542 if (icmplen < ICMP_TSLEN) {
543 ICMPSTAT_INC(icps_badlen);
546 icp->icmp_type = ICMP_TSTAMPREPLY;
547 icp->icmp_rtime = iptime();
548 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
549 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
555 if (V_icmpmaskrepl == 0)
558 * We are not able to respond with all ones broadcast
559 * unless we receive it over a point-to-point interface.
561 if (icmplen < ICMP_MASKLEN)
563 switch (ip->ip_dst.s_addr) {
565 case INADDR_BROADCAST:
567 icmpdst.sin_addr = ip->ip_src;
571 icmpdst.sin_addr = ip->ip_dst;
573 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
574 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
577 if (ia->ia_ifp == NULL) {
578 ifa_free(&ia->ia_ifa);
581 icp->icmp_type = ICMP_MASKREPLY;
582 if (V_icmpmaskfake == 0)
583 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
585 icp->icmp_mask = V_icmpmaskfake;
586 if (ip->ip_src.s_addr == 0) {
587 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
588 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
589 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
590 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
592 ifa_free(&ia->ia_ifa);
594 ip->ip_len += hlen; /* since ip_input deducts this */
595 ICMPSTAT_INC(icps_reflect);
596 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
601 if (V_log_redirect) {
604 src = ntohl(ip->ip_src.s_addr);
605 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
606 gw = ntohl(icp->icmp_gwaddr.s_addr);
607 printf("icmp redirect from %d.%d.%d.%d: "
608 "%d.%d.%d.%d => %d.%d.%d.%d\n",
609 (int)(src >> 24), (int)((src >> 16) & 0xff),
610 (int)((src >> 8) & 0xff), (int)(src & 0xff),
611 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
612 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
613 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
614 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
617 * RFC1812 says we must ignore ICMP redirects if we
618 * are acting as router.
620 if (V_drop_redirect || V_ipforwarding)
624 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
625 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
626 ICMPSTAT_INC(icps_badlen);
630 * Short circuit routing redirects to force
631 * immediate change in the kernel's routing
632 * tables. The message is also handed to anyone
633 * listening on a raw socket (e.g. the routing
634 * daemon for use in updating its tables).
636 icmpgw.sin_addr = ip->ip_src;
637 icmpdst.sin_addr = icp->icmp_gwaddr;
640 char buf[4 * sizeof "123"];
641 strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
643 printf("redirect dst %s to %s\n",
644 buf, inet_ntoa(icp->icmp_gwaddr));
647 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
648 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
649 in_rtredirect((struct sockaddr *)&icmpsrc,
650 (struct sockaddr *)&icmpdst,
651 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
652 (struct sockaddr *)&icmpgw, fibnum);
654 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
656 key_sa_routechange((struct sockaddr *)&icmpsrc);
661 * No kernel processing for the following;
662 * just fall through to send to raw listener.
665 case ICMP_ROUTERADVERT:
666 case ICMP_ROUTERSOLICIT:
667 case ICMP_TSTAMPREPLY:
683 * Reflect the ip packet back to the source
686 icmp_reflect(struct mbuf *m)
688 struct ip *ip = mtod(m, struct ip *);
691 struct in_ifaddr *ia;
693 struct mbuf *opts = 0;
694 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
696 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
697 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
698 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
699 m_freem(m); /* Bad return address */
700 ICMPSTAT_INC(icps_badaddr);
701 goto done; /* Ip_output() will check for broadcast */
705 ip->ip_dst = ip->ip_src;
708 * Source selection for ICMP replies:
710 * If the incoming packet was addressed directly to one of our
711 * own addresses, use dst as the src for the reply.
714 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
715 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
716 t = IA_SIN(ia)->sin_addr;
724 * If the incoming packet was addressed to one of our broadcast
725 * addresses, use the first non-broadcast address which corresponds
726 * to the incoming interface.
728 ifp = m->m_pkthdr.rcvif;
729 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
731 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
732 if (ifa->ifa_addr->sa_family != AF_INET)
735 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
737 t = IA_SIN(ia)->sin_addr;
738 IF_ADDR_RUNLOCK(ifp);
742 IF_ADDR_RUNLOCK(ifp);
745 * If the packet was transiting through us, use the address of
746 * the interface the packet came through in. If that interface
747 * doesn't have a suitable IP address, the normal selection
750 if (V_icmp_rfi && ifp != NULL) {
752 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
753 if (ifa->ifa_addr->sa_family != AF_INET)
756 t = IA_SIN(ia)->sin_addr;
757 IF_ADDR_RUNLOCK(ifp);
760 IF_ADDR_RUNLOCK(ifp);
763 * If the incoming packet was not addressed directly to us, use
764 * designated interface for icmp replies specified by sysctl
765 * net.inet.icmp.reply_src (default not set). Otherwise continue
766 * with normal source selection.
768 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
770 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
771 if (ifa->ifa_addr->sa_family != AF_INET)
774 t = IA_SIN(ia)->sin_addr;
775 IF_ADDR_RUNLOCK(ifp);
778 IF_ADDR_RUNLOCK(ifp);
781 * If the packet was transiting through us, use the address of
782 * the interface that is the closest to the packet source.
783 * When we don't have a route back to the packet source, stop here
784 * and drop the packet.
786 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
789 ICMPSTAT_INC(icps_noroute);
792 t = IA_SIN(ia)->sin_addr;
793 ifa_free(&ia->ia_ifa);
796 mac_netinet_icmp_replyinplace(m);
799 ip->ip_ttl = V_ip_defttl;
807 * Retrieve any source routing from the incoming packet;
808 * add on any record-route or timestamp options.
810 cp = (u_char *) (ip + 1);
811 if ((opts = ip_srcroute(m)) == 0 &&
812 (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
813 opts->m_len = sizeof(struct in_addr);
814 mtod(opts, struct in_addr *)->s_addr = 0;
819 printf("icmp_reflect optlen %d rt %d => ",
820 optlen, opts->m_len);
822 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
823 opt = cp[IPOPT_OPTVAL];
824 if (opt == IPOPT_EOL)
826 if (opt == IPOPT_NOP)
829 if (cnt < IPOPT_OLEN + sizeof(*cp))
831 len = cp[IPOPT_OLEN];
832 if (len < IPOPT_OLEN + sizeof(*cp) ||
837 * Should check for overflow, but it "can't happen"
839 if (opt == IPOPT_RR || opt == IPOPT_TS ||
840 opt == IPOPT_SECURITY) {
842 mtod(opts, caddr_t) + opts->m_len, len);
846 /* Terminate & pad, if necessary */
847 cnt = opts->m_len % 4;
849 for (; cnt < 4; cnt++) {
850 *(mtod(opts, caddr_t) + opts->m_len) =
857 printf("%d\n", opts->m_len);
861 * Now strip out original options by copying rest of first
862 * mbuf's data back, and adjust the IP length.
864 ip->ip_len -= optlen;
865 ip->ip_v = IPVERSION;
868 if (m->m_flags & M_PKTHDR)
869 m->m_pkthdr.len -= optlen;
870 optlen += sizeof(struct ip);
871 bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
872 (unsigned)(m->m_len - sizeof(struct ip)));
874 m_tag_delete_nonpersistent(m);
875 m->m_flags &= ~(M_BCAST|M_MCAST);
883 * Send an icmp packet back to the ip level,
884 * after supplying a checksum.
887 icmp_send(struct mbuf *m, struct mbuf *opts)
889 register struct ip *ip = mtod(m, struct ip *);
891 register struct icmp *icp;
893 hlen = ip->ip_hl << 2;
896 icp = mtod(m, struct icmp *);
898 icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
901 m->m_pkthdr.rcvif = (struct ifnet *)0;
904 char buf[4 * sizeof "123"];
905 strcpy(buf, inet_ntoa(ip->ip_dst));
906 printf("icmp_send dst %s src %s\n",
907 buf, inet_ntoa(ip->ip_src));
910 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
914 * Return milliseconds since 00:00 GMT in network format.
923 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
928 * Return the next larger or smaller MTU plateau (table from RFC 1191)
929 * given current value MTU. If DIR is less than zero, a larger plateau
930 * is returned; otherwise, a smaller value is returned.
933 ip_next_mtu(int mtu, int dir)
935 static int mtutab[] = {
936 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
941 size = (sizeof mtutab) / (sizeof mtutab[0]);
943 for (i = 0; i < size; i++)
947 for (i = size - 1; i >= 0; i--)
950 if (mtu == mtutab[0])
958 * badport_bandlim() - check for ICMP bandwidth limit
960 * Return 0 if it is ok to send an ICMP error response, -1 if we have
961 * hit our bandwidth limit and it is not ok.
963 * If icmplim is <= 0, the feature is disabled and 0 is returned.
965 * For now we separate the TCP and UDP subsystems w/ different 'which'
966 * values. We may eventually remove this separation (and simplify the
969 * Note that the printing of the error message is delayed so we can
970 * properly print the icmp error rate that the system was trying to do
971 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
972 * the 'final' error, but it doesn't make sense to solve the printing
973 * delay with more complex code.
977 badport_bandlim(int which)
980 #define N(a) (sizeof (a) / sizeof (a[0]))
983 struct timeval lasttime;
985 } rates[BANDLIM_MAX+1] = {
986 { "icmp unreach response" },
987 { "icmp ping response" },
988 { "icmp tstamp response" },
989 { "closed port RST response" },
990 { "open port RST response" },
991 { "icmp6 unreach response" },
992 { "sctp ootb response" }
996 * Return ok status if feature disabled or argument out of range.
998 if (V_icmplim > 0 && (u_int) which < N(rates)) {
999 struct rate *r = &rates[which];
1000 int opps = r->curpps;
1002 if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
1003 return -1; /* discard packet */
1005 * If we've dropped below the threshold after having
1006 * rate-limited traffic print the message. This preserves
1007 * the previous behaviour at the expense of added complexity.
1009 if (V_icmplim_output && opps > V_icmplim)
1010 log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1011 r->type, opps, V_icmplim);
1013 return 0; /* okay to send packet */