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 * 3. 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_RW,
155 &VNET_NAME(icmptstamprepl), 0,
156 "Respond to ICMP Timestamp packets");
162 static void icmp_reflect(struct mbuf *);
163 static void icmp_send(struct mbuf *, struct mbuf *);
165 extern struct protosw inetsw[];
168 * Kernel module interface for updating icmpstat. The argument is an index
169 * into icmpstat treated as an array of u_long. While this encodes the
170 * general layout of icmpstat into the caller, it doesn't encode its
171 * location, so that future changes to add, for example, per-CPU stats
172 * support won't cause binary compatibility problems for kernel modules.
175 kmod_icmpstat_inc(int statnum)
178 counter_u64_add(VNET(icmpstat)[statnum], 1);
182 * Generate an error packet of type error
183 * in response to bad packet ip.
186 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
188 struct ip *oip, *nip;
191 unsigned icmplen, icmpelen, nlen, oiphlen;
193 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
196 if (type != ICMP_REDIRECT)
197 ICMPSTAT_INC(icps_error);
200 * if the original packet was encrypted.
201 * if not the first fragment of message.
202 * in response to a multicast or broadcast packet.
203 * if the old packet protocol was an ICMP error message.
205 if (n->m_flags & M_DECRYPTED)
207 if (n->m_flags & (M_BCAST|M_MCAST))
210 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
211 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
213 oip = mtod(n, struct ip *);
214 oiphlen = oip->ip_hl << 2;
215 if (n->m_len < oiphlen + ICMP_MINLEN)
219 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
221 if (oip->ip_off & htons(~(IP_MF|IP_DF)))
223 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
224 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
225 oiphlen))->icmp_type)) {
226 ICMPSTAT_INC(icps_oldicmp);
230 * Calculate length to quote from original packet and
231 * prevent the ICMP mbuf from overflowing.
232 * Unfortunately this is non-trivial since ip_forward()
233 * sends us truncated packets.
235 nlen = m_length(n, NULL);
236 if (oip->ip_p == IPPROTO_TCP) {
240 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
243 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
244 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
246 oip = mtod(n, struct ip *);
247 th = mtodo(n, oiphlen);
248 tcphlen = th->th_off << 2;
249 if (tcphlen < sizeof(struct tcphdr))
251 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
253 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
255 if (n->m_len < oiphlen + tcphlen &&
256 (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
258 icmpelen = max(tcphlen, min(V_icmp_quotelen,
259 ntohs(oip->ip_len) - oiphlen));
260 } else if (oip->ip_p == IPPROTO_SCTP) {
262 struct sctp_chunkhdr *ch;
264 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
266 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
269 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
270 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
272 oip = mtod(n, struct ip *);
273 icmpelen = max(sizeof(struct sctphdr),
274 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
275 sh = mtodo(n, oiphlen);
276 if (ntohl(sh->v_tag) == 0 &&
277 ntohs(oip->ip_len) >= oiphlen +
278 sizeof(struct sctphdr) + 8 &&
279 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
280 n->m_next != NULL)) {
281 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
282 (n = m_pullup(n, oiphlen +
283 sizeof(struct sctphdr) + 8)) == NULL)
285 oip = mtod(n, struct ip *);
286 sh = mtodo(n, oiphlen);
287 ch = (struct sctp_chunkhdr *)(sh + 1);
288 if (ch->chunk_type == SCTP_INITIATION) {
289 icmpelen = max(sizeof(struct sctphdr) + 8,
290 min(V_icmp_quotelen, ntohs(oip->ip_len) -
295 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
298 icmplen = min(oiphlen + icmpelen, nlen);
299 if (icmplen < sizeof(struct ip))
302 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
303 m = m_gethdr(M_NOWAIT, MT_DATA);
305 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
309 mac_netinet_icmp_reply(n, m);
311 icmplen = min(icmplen, M_TRAILINGSPACE(m) -
312 sizeof(struct ip) - ICMP_MINLEN);
313 m_align(m, ICMP_MINLEN + icmplen);
314 m->m_len = ICMP_MINLEN + icmplen;
316 /* XXX MRT make the outgoing packet use the same FIB
317 * that was associated with the incoming packet
319 M_SETFIB(m, M_GETFIB(n));
320 icp = mtod(m, struct icmp *);
321 ICMPSTAT_INC(icps_outhist[type]);
322 icp->icmp_type = type;
323 if (type == ICMP_REDIRECT)
324 icp->icmp_gwaddr.s_addr = dest;
328 * The following assignments assume an overlay with the
329 * just zeroed icmp_void field.
331 if (type == ICMP_PARAMPROB) {
332 icp->icmp_pptr = code;
334 } else if (type == ICMP_UNREACH &&
335 code == ICMP_UNREACH_NEEDFRAG && mtu) {
336 icp->icmp_nextmtu = htons(mtu);
339 icp->icmp_code = code;
342 * Copy the quotation into ICMP message and
343 * convert quoted IP header back to network representation.
345 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
349 * Set up ICMP message mbuf and copy old IP header (without options
350 * in front of ICMP message.
351 * If the original mbuf was meant to bypass the firewall, the error
352 * reply should bypass as well.
354 m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
355 m->m_data -= sizeof(struct ip);
356 m->m_len += sizeof(struct ip);
357 m->m_pkthdr.len = m->m_len;
358 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
359 nip = mtod(m, struct ip *);
360 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
361 nip->ip_len = htons(m->m_len);
362 nip->ip_v = IPVERSION;
364 nip->ip_p = IPPROTO_ICMP;
374 * Process a received ICMP message.
377 icmp_input(struct mbuf **mp, int *offp, int proto)
380 struct in_ifaddr *ia;
381 struct mbuf *m = *mp;
382 struct ip *ip = mtod(m, struct ip *);
383 struct sockaddr_in icmpsrc, icmpdst, icmpgw;
385 int icmplen = ntohs(ip->ip_len) - *offp;
387 void (*ctlfunc)(int, struct sockaddr *, void *);
393 * Locate icmp structure in mbuf, and check
394 * that not corrupted and of at least minimum length.
398 char srcbuf[INET_ADDRSTRLEN];
399 char dstbuf[INET_ADDRSTRLEN];
401 printf("icmp_input from %s to %s, len %d\n",
402 inet_ntoa_r(ip->ip_src, srcbuf),
403 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
406 if (icmplen < ICMP_MINLEN) {
407 ICMPSTAT_INC(icps_tooshort);
410 i = hlen + min(icmplen, ICMP_ADVLENMIN);
411 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
412 ICMPSTAT_INC(icps_tooshort);
413 return (IPPROTO_DONE);
415 ip = mtod(m, struct ip *);
418 icp = mtod(m, struct icmp *);
419 if (in_cksum(m, icmplen)) {
420 ICMPSTAT_INC(icps_checksum);
428 printf("icmp_input, type %d code %d\n", icp->icmp_type,
433 * Message type specific processing.
435 if (icp->icmp_type > ICMP_MAXTYPE)
439 bzero(&icmpsrc, sizeof(icmpsrc));
440 icmpsrc.sin_len = sizeof(struct sockaddr_in);
441 icmpsrc.sin_family = AF_INET;
442 bzero(&icmpdst, sizeof(icmpdst));
443 icmpdst.sin_len = sizeof(struct sockaddr_in);
444 icmpdst.sin_family = AF_INET;
445 bzero(&icmpgw, sizeof(icmpgw));
446 icmpgw.sin_len = sizeof(struct sockaddr_in);
447 icmpgw.sin_family = AF_INET;
449 ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
450 code = icp->icmp_code;
451 switch (icp->icmp_type) {
455 case ICMP_UNREACH_NET:
456 case ICMP_UNREACH_HOST:
457 case ICMP_UNREACH_SRCFAIL:
458 case ICMP_UNREACH_NET_UNKNOWN:
459 case ICMP_UNREACH_HOST_UNKNOWN:
460 case ICMP_UNREACH_ISOLATED:
461 case ICMP_UNREACH_TOSNET:
462 case ICMP_UNREACH_TOSHOST:
463 case ICMP_UNREACH_HOST_PRECEDENCE:
464 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
465 code = PRC_UNREACH_NET;
468 case ICMP_UNREACH_NEEDFRAG:
473 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
474 * Treat subcodes 2,3 as immediate RST
476 case ICMP_UNREACH_PROTOCOL:
477 code = PRC_UNREACH_PROTOCOL;
479 case ICMP_UNREACH_PORT:
480 code = PRC_UNREACH_PORT;
483 case ICMP_UNREACH_NET_PROHIB:
484 case ICMP_UNREACH_HOST_PROHIB:
485 case ICMP_UNREACH_FILTER_PROHIB:
486 code = PRC_UNREACH_ADMIN_PROHIB;
497 code += PRC_TIMXCEED_INTRANS;
503 code = PRC_PARAMPROB;
506 * Problem with datagram; advise higher level routines.
508 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
509 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
510 ICMPSTAT_INC(icps_badlen);
513 /* Discard ICMP's in response to multicast packets */
514 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
518 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
520 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
522 * XXX if the packet contains [IPv4 AH TCP], we can't make a
523 * notification to TCP layer.
525 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
527 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
528 /* This should actually not happen */
529 ICMPSTAT_INC(icps_tooshort);
530 return (IPPROTO_DONE);
532 ip = mtod(m, struct ip *);
533 icp = (struct icmp *)(ip + 1);
535 * The upper layer handler can rely on:
536 * - The outer IP header has no options.
537 * - The outer IP header, the ICMP header, the inner IP header,
538 * and the first n bytes of the inner payload are contiguous.
539 * n is at least 8, but might be larger based on
540 * ICMP_ADVLENPREF. See its definition in ip_icmp.h.
542 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
544 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
545 (void *)&icp->icmp_ip);
549 ICMPSTAT_INC(icps_badcode);
553 if (!V_icmpbmcastecho
554 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
555 ICMPSTAT_INC(icps_bmcastecho);
558 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
560 icp->icmp_type = ICMP_ECHOREPLY;
564 if (V_icmptstamprepl == 0)
566 if (!V_icmpbmcastecho
567 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
568 ICMPSTAT_INC(icps_bmcasttstamp);
571 if (icmplen < ICMP_TSLEN) {
572 ICMPSTAT_INC(icps_badlen);
575 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
577 icp->icmp_type = ICMP_TSTAMPREPLY;
578 icp->icmp_rtime = iptime();
579 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
583 if (V_icmpmaskrepl == 0)
586 * We are not able to respond with all ones broadcast
587 * unless we receive it over a point-to-point interface.
589 if (icmplen < ICMP_MASKLEN)
591 switch (ip->ip_dst.s_addr) {
593 case INADDR_BROADCAST:
595 icmpdst.sin_addr = ip->ip_src;
599 icmpdst.sin_addr = ip->ip_dst;
601 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
602 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
605 if (ia->ia_ifp == NULL) {
606 ifa_free(&ia->ia_ifa);
609 icp->icmp_type = ICMP_MASKREPLY;
610 if (V_icmpmaskfake == 0)
611 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
613 icp->icmp_mask = V_icmpmaskfake;
614 if (ip->ip_src.s_addr == 0) {
615 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
616 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
617 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
618 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
620 ifa_free(&ia->ia_ifa);
622 ICMPSTAT_INC(icps_reflect);
623 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
625 return (IPPROTO_DONE);
628 if (V_log_redirect) {
631 src = ntohl(ip->ip_src.s_addr);
632 dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
633 gw = ntohl(icp->icmp_gwaddr.s_addr);
634 printf("icmp redirect from %d.%d.%d.%d: "
635 "%d.%d.%d.%d => %d.%d.%d.%d\n",
636 (int)(src >> 24), (int)((src >> 16) & 0xff),
637 (int)((src >> 8) & 0xff), (int)(src & 0xff),
638 (int)(dst >> 24), (int)((dst >> 16) & 0xff),
639 (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
640 (int)(gw >> 24), (int)((gw >> 16) & 0xff),
641 (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
644 * RFC1812 says we must ignore ICMP redirects if we
645 * are acting as router.
647 if (V_drop_redirect || V_ipforwarding)
651 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
652 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
653 ICMPSTAT_INC(icps_badlen);
657 * Short circuit routing redirects to force
658 * immediate change in the kernel's routing
659 * tables. The message is also handed to anyone
660 * listening on a raw socket (e.g. the routing
661 * daemon for use in updating its tables).
663 icmpgw.sin_addr = ip->ip_src;
664 icmpdst.sin_addr = icp->icmp_gwaddr;
667 char dstbuf[INET_ADDRSTRLEN];
668 char gwbuf[INET_ADDRSTRLEN];
670 printf("redirect dst %s to %s\n",
671 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
672 inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
675 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
676 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
677 in_rtredirect((struct sockaddr *)&icmpsrc,
678 (struct sockaddr *)&icmpdst,
679 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
680 (struct sockaddr *)&icmpgw, fibnum);
682 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
686 * No kernel processing for the following;
687 * just fall through to send to raw listener.
690 case ICMP_ROUTERADVERT:
691 case ICMP_ROUTERSOLICIT:
692 case ICMP_TSTAMPREPLY:
695 case ICMP_SOURCEQUENCH:
702 rip_input(mp, offp, proto);
703 return (IPPROTO_DONE);
707 return (IPPROTO_DONE);
711 * Reflect the ip packet back to the source
714 icmp_reflect(struct mbuf *m)
716 struct rm_priotracker in_ifa_tracker;
717 struct ip *ip = mtod(m, struct ip *);
720 struct in_ifaddr *ia;
722 struct nhop4_extended nh_ext;
723 struct mbuf *opts = NULL;
724 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
726 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
727 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
728 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
729 m_freem(m); /* Bad return address */
730 ICMPSTAT_INC(icps_badaddr);
731 goto done; /* Ip_output() will check for broadcast */
735 ip->ip_dst = ip->ip_src;
738 * Source selection for ICMP replies:
740 * If the incoming packet was addressed directly to one of our
741 * own addresses, use dst as the src for the reply.
743 IN_IFADDR_RLOCK(&in_ifa_tracker);
744 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
745 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
746 t = IA_SIN(ia)->sin_addr;
747 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
751 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
754 * If the incoming packet was addressed to one of our broadcast
755 * addresses, use the first non-broadcast address which corresponds
756 * to the incoming interface.
758 ifp = m->m_pkthdr.rcvif;
759 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
761 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
762 if (ifa->ifa_addr->sa_family != AF_INET)
765 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
767 t = IA_SIN(ia)->sin_addr;
768 IF_ADDR_RUNLOCK(ifp);
772 IF_ADDR_RUNLOCK(ifp);
775 * If the packet was transiting through us, use the address of
776 * the interface the packet came through in. If that interface
777 * doesn't have a suitable IP address, the normal selection
780 if (V_icmp_rfi && ifp != NULL) {
782 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
783 if (ifa->ifa_addr->sa_family != AF_INET)
786 t = IA_SIN(ia)->sin_addr;
787 IF_ADDR_RUNLOCK(ifp);
790 IF_ADDR_RUNLOCK(ifp);
793 * If the incoming packet was not addressed directly to us, use
794 * designated interface for icmp replies specified by sysctl
795 * net.inet.icmp.reply_src (default not set). Otherwise continue
796 * with normal source selection.
798 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
800 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
801 if (ifa->ifa_addr->sa_family != AF_INET)
804 t = IA_SIN(ia)->sin_addr;
805 IF_ADDR_RUNLOCK(ifp);
808 IF_ADDR_RUNLOCK(ifp);
811 * If the packet was transiting through us, use the address of
812 * the interface that is the closest to the packet source.
813 * When we don't have a route back to the packet source, stop here
814 * and drop the packet.
816 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
818 ICMPSTAT_INC(icps_noroute);
824 mac_netinet_icmp_replyinplace(m);
827 ip->ip_ttl = V_ip_defttl;
835 * Retrieve any source routing from the incoming packet;
836 * add on any record-route or timestamp options.
838 cp = (u_char *) (ip + 1);
839 if ((opts = ip_srcroute(m)) == NULL &&
840 (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
841 opts->m_len = sizeof(struct in_addr);
842 mtod(opts, struct in_addr *)->s_addr = 0;
847 printf("icmp_reflect optlen %d rt %d => ",
848 optlen, opts->m_len);
850 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
851 opt = cp[IPOPT_OPTVAL];
852 if (opt == IPOPT_EOL)
854 if (opt == IPOPT_NOP)
857 if (cnt < IPOPT_OLEN + sizeof(*cp))
859 len = cp[IPOPT_OLEN];
860 if (len < IPOPT_OLEN + sizeof(*cp) ||
865 * Should check for overflow, but it "can't happen"
867 if (opt == IPOPT_RR || opt == IPOPT_TS ||
868 opt == IPOPT_SECURITY) {
870 mtod(opts, caddr_t) + opts->m_len, len);
874 /* Terminate & pad, if necessary */
875 cnt = opts->m_len % 4;
877 for (; cnt < 4; cnt++) {
878 *(mtod(opts, caddr_t) + opts->m_len) =
885 printf("%d\n", opts->m_len);
890 m_tag_delete_nonpersistent(m);
891 m->m_flags &= ~(M_BCAST|M_MCAST);
899 * Send an icmp packet back to the ip level,
900 * after supplying a checksum.
903 icmp_send(struct mbuf *m, struct mbuf *opts)
905 struct ip *ip = mtod(m, struct ip *);
909 hlen = ip->ip_hl << 2;
912 icp = mtod(m, struct icmp *);
914 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
917 m->m_pkthdr.rcvif = (struct ifnet *)0;
920 char dstbuf[INET_ADDRSTRLEN];
921 char srcbuf[INET_ADDRSTRLEN];
923 printf("icmp_send dst %s src %s\n",
924 inet_ntoa_r(ip->ip_dst, dstbuf),
925 inet_ntoa_r(ip->ip_src, srcbuf));
928 (void) ip_output(m, opts, NULL, 0, NULL, NULL);
932 * Return milliseconds since 00:00 UTC in network format.
941 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
946 * Return the next larger or smaller MTU plateau (table from RFC 1191)
947 * given current value MTU. If DIR is less than zero, a larger plateau
948 * is returned; otherwise, a smaller value is returned.
951 ip_next_mtu(int mtu, int dir)
953 static int mtutab[] = {
954 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
959 size = (sizeof mtutab) / (sizeof mtutab[0]);
961 for (i = 0; i < size; i++)
965 for (i = size - 1; i >= 0; i--)
968 if (mtu == mtutab[0])
977 * badport_bandlim() - check for ICMP bandwidth limit
979 * Return 0 if it is ok to send an ICMP error response, -1 if we have
980 * hit our bandwidth limit and it is not ok.
982 * If icmplim is <= 0, the feature is disabled and 0 is returned.
984 * For now we separate the TCP and UDP subsystems w/ different 'which'
985 * values. We may eventually remove this separation (and simplify the
988 * Note that the printing of the error message is delayed so we can
989 * properly print the icmp error rate that the system was trying to do
990 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
991 * the 'final' error, but it doesn't make sense to solve the printing
992 * delay with more complex code.
996 struct counter_rate cr;
998 static VNET_DEFINE(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
999 { "icmp unreach response" },
1000 { "icmp ping response" },
1001 { "icmp tstamp response" },
1002 { "closed port RST response" },
1003 { "open port RST response" },
1004 { "icmp6 unreach response" },
1005 { "sctp ootb response" }
1007 #define V_icmp_rates VNET(icmp_rates)
1010 icmp_bandlimit_init(void)
1013 for (int i = 0; i < BANDLIM_MAX; i++) {
1014 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1015 V_icmp_rates[i].cr.cr_ticks = ticks;
1018 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1019 icmp_bandlimit_init, NULL);
1022 icmp_bandlimit_uninit(void)
1025 for (int i = 0; i < BANDLIM_MAX; i++)
1026 counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1028 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1029 icmp_bandlimit_uninit, NULL);
1032 badport_bandlim(int which)
1036 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1039 KASSERT(which >= 0 && which < BANDLIM_MAX,
1040 ("%s: which %d", __func__, which));
1042 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim);
1045 if (pps > 0 && V_icmplim_output)
1046 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1047 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim);