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
30 * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
33 #include "opt_ipsec.h"
35 #include <sys/param.h>
36 #include <sys/systm.h>
38 #include <sys/protosw.h>
39 #include <sys/socket.h>
40 #include <sys/socketops.h>
42 #include <sys/kernel.h>
43 #include <sys/sysctl.h>
44 #include <sys/in_cksum.h>
46 #include <machine/stdarg.h>
49 #include <net/if_types.h>
50 #include <net/netisr2.h>
51 #include <net/netmsg2.h>
52 #include <net/route.h>
55 #include <netinet/in.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/in_var.h>
58 #include <netinet/ip.h>
59 #include <netinet/ip_icmp.h>
60 #include <netinet/ip_var.h>
61 #include <netinet/icmp_var.h>
64 #include <netinet6/ipsec.h>
65 #include <netproto/key/key.h>
69 #include <netproto/ipsec/ipsec.h>
70 #include <netproto/ipsec/key.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.
80 struct icmpstat icmpstat;
81 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
82 &icmpstat, icmpstat, "ICMP statistics");
84 static int icmpmaskrepl = 0;
85 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
86 &icmpmaskrepl, 0, "Allow replies to netmask requests");
88 static int drop_redirect = 0;
89 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
90 &drop_redirect, 0, "Ignore ICMP redirects");
92 static int log_redirect = 0;
93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
94 &log_redirect, 0, "Enable output about ICMP redirects");
96 static int discard_sourcequench = 1;
97 SYSCTL_INT(_net_inet_icmp, OID_AUTO, discard_sourcequench, CTLFLAG_RW,
98 &discard_sourcequench, 0, "Discard ICMP Source Quench");
103 * ICMP error-response bandwidth limiting sysctl. If not enabled, sysctl
104 * variable content is -1 and read-only.
107 static int icmplim = 200;
108 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
109 &icmplim, 0, "ICMP bandwidth limit");
112 static int icmplim = -1;
113 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
114 &icmplim, 0, "ICMP bandwidth limit");
118 static int icmplim_output = 0;
119 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
120 &icmplim_output, 0, "Enable output about ICMP bandwidth limits");
123 * ICMP broadcast echo sysctl
126 static int icmpbmcastecho = 0;
127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
128 &icmpbmcastecho, 0, "");
130 static char icmp_reply_src[IFNAMSIZ];
131 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
132 icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
135 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
136 &icmp_rfi, 0, "ICMP reply from incoming interface for "
137 "non-local packets");
140 static int icmpprintfs = 0;
141 SYSCTL_INT(_net_inet_icmp, OID_AUTO, debug_prints, CTLFLAG_RW,
142 &icmpprintfs, 0, "extra ICMP debug prints");
145 static void icmp_reflect (struct mbuf *);
146 static void icmp_send (struct mbuf *, struct mbuf *, struct route *);
148 extern struct protosw inetsw[];
151 * Generate an error packet of type error
152 * in response to bad packet ip.
155 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
157 struct ip *oip = mtod(n, struct ip *), *nip;
158 unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
165 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
167 if (type != ICMP_REDIRECT)
168 icmpstat.icps_error++;
170 * Don't send error if the original packet was encrypted.
171 * Don't send error if not the first fragment of message.
172 * Don't error if the old packet protocol was ICMP
173 * error message, only known informational types.
175 if (n->m_flags & M_DECRYPTED)
177 if (oip->ip_off &~ (IP_MF|IP_DF))
179 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
180 n->m_len >= oiplen + ICMP_MINLEN &&
181 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
182 icmpstat.icps_oldicmp++;
185 /* Don't send error in response to a multicast or broadcast packet */
186 if (n->m_flags & (M_BCAST|M_MCAST))
189 * First, formulate icmp message
191 m = m_gethdr(M_NOWAIT, MT_HEADER);
194 icmplen = min(oiplen + 8, oip->ip_len);
195 if (icmplen < sizeof(struct ip))
196 panic("icmp_error: bad length");
197 m->m_len = icmplen + ICMP_MINLEN;
198 MH_ALIGN(m, m->m_len);
199 icp = mtod(m, struct icmp *);
200 if ((u_int)type > ICMP_MAXTYPE)
202 icmpstat.icps_outhist[type]++;
203 icp->icmp_type = type;
204 if (type == ICMP_REDIRECT)
205 icp->icmp_gwaddr.s_addr = dest;
209 * The following assignments assume an overlay with the
210 * zeroed icmp_void field.
212 if (type == ICMP_PARAMPROB) {
213 icp->icmp_pptr = code;
215 } else if (type == ICMP_UNREACH &&
216 code == ICMP_UNREACH_NEEDFRAG && destmtu) {
217 icp->icmp_nextmtu = htons(destmtu);
221 icp->icmp_code = code;
222 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
226 * Convert fields to network representation.
228 nip->ip_len = htons(nip->ip_len);
229 nip->ip_off = htons(nip->ip_off);
232 * Now, copy old ip header (without options)
233 * in front of icmp message.
235 if (m->m_data - sizeof(struct ip) < m->m_pktdat)
237 m->m_data -= sizeof(struct ip);
238 m->m_len += sizeof(struct ip);
239 m->m_pkthdr.len = m->m_len;
240 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
241 nip = mtod(m, struct ip *);
242 bcopy(oip, nip, sizeof(struct ip));
243 nip->ip_len = m->m_len;
244 nip->ip_vhl = IP_VHL_BORING;
245 nip->ip_p = IPPROTO_ICMP;
247 m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
255 icmp_ctlinput_done_handler(netmsg_t nmsg)
257 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
258 struct mbuf *m = msg->m;
259 int hlen = msg->hlen;
261 rip_input(&m, &hlen, msg->proto);
265 icmp_ctlinput_done(struct mbuf *m)
267 struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg;
269 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
270 icmp_ctlinput_done_handler);
271 lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg);
275 icmp_mtudisc(struct mbuf *m, int hlen)
277 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
281 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
283 icp = mtodoff(m, struct icmp *, hlen);
284 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
288 * If we got a needfrag and there is a host route to the original
289 * destination, and the MTU is not locked, then set the MTU in the
290 * route to the suggested new value (if given) and then notify as
291 * usual. The ULPs will notice that the MTU has changed and adapt
292 * accordingly. If no new MTU was suggested, then we guess a new
293 * one less than the current value. If the new MTU is unreasonably
294 * small (arbitrarily set at 296), then we reset the MTU to the
295 * interface value and enable the lock bit, indicating that we are
296 * no longer doing MTU discovery.
298 rt = rtpurelookup((struct sockaddr *)&icmpsrc);
299 if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
300 !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
302 char src_buf[INET_ADDRSTRLEN];
306 mtu = ntohs(icp->icmp_nextmtu);
308 mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu, 1);
310 kprintf("MTU for %s reduced to %d\n",
311 inet_ntop(AF_INET, &icmpsrc.sin_addr,
312 src_buf, INET_ADDRSTRLEN), mtu);
315 /* rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu; */
316 rt->rt_rmx.rmx_locks |= RTV_MTU;
317 } else if (rt->rt_rmx.rmx_mtu > mtu) {
318 rt->rt_rmx.rmx_mtu = mtu;
325 * XXX if the packet contains [IPv4 AH TCP], we can't make a
326 * notification to TCP layer.
328 so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]],
329 PRC_MSGSIZE, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
333 icmp_mtudisc_handler(netmsg_t nmsg)
335 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
338 icmp_mtudisc(msg->m, msg->hlen);
340 nextcpu = mycpuid + 1;
342 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg);
344 icmp_ctlinput_done(msg->m);
348 icmp_mtudisc_start(struct mbuf *m, int hlen, int proto)
350 struct netmsg_ctlinput *msg;
352 KASSERT(&curthread->td_msgport == netisr_cpuport(0),
355 icmp_mtudisc(m, hlen);
358 /* There is only one netisr; done */
362 msg = &m->m_hdr.mh_ctlmsg;
363 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
364 icmp_mtudisc_handler);
366 msg->cmd = PRC_MSGSIZE;
370 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
375 icmp_ctlinput(struct mbuf *m, int cmd, int hlen)
377 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
380 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
382 icp = mtodoff(m, struct icmp *, hlen);
383 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
386 * XXX if the packet contains [IPv4 AH TCP], we can't make a
387 * notification to TCP layer.
389 so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]],
390 cmd, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
394 icmp_ctlinput_handler(netmsg_t nmsg)
396 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
398 icmp_ctlinput(msg->m, msg->cmd, msg->hlen);
399 icmp_ctlinput_done(msg->m);
403 icmp_ctlinput_start(struct mbuf *m, struct lwkt_port *port,
404 int cmd, int hlen, int proto)
406 struct netmsg_ctlinput *msg;
408 KASSERT(&curthread->td_msgport != port,
409 ("send icmp ctlinput to the current netisr"));
411 msg = &m->m_hdr.mh_ctlmsg;
412 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
413 icmp_ctlinput_handler);
419 lwkt_sendmsg(port, &msg->base.lmsg);
423 icmp_ctlinput_global_handler(netmsg_t nmsg)
425 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
428 icmp_ctlinput(msg->m, msg->cmd, msg->hlen);
430 nextcpu = mycpuid + 1;
432 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg);
434 icmp_ctlinput_done(msg->m);
438 icmp_ctlinput_global_start(struct mbuf *m, int cmd, int hlen, int proto)
440 struct netmsg_ctlinput *msg;
442 KASSERT(&curthread->td_msgport == netisr_cpuport(0),
444 KASSERT(ncpus > 1, ("there is only 1 cpu"));
446 icmp_ctlinput(m, cmd, hlen);
448 msg = &m->m_hdr.mh_ctlmsg;
449 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
450 icmp_ctlinput_global_handler);
456 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
459 #define ICMP_RTREDIRECT_FLAGS (RTF_GATEWAY | RTF_HOST)
462 icmp_redirect(struct mbuf *m, int hlen, boolean_t prt)
464 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
465 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
466 struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
470 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
472 ip = mtod(m, struct ip *);
473 icp = mtodoff(m, struct icmp *, hlen);
476 * Short circuit routing redirects to force immediate change
477 * in the kernel's routing tables. The message is also handed
478 * to anyone listening on a raw socket (e.g. the routing daemon
479 * for use in updating its tables).
482 if (icmpprintfs && prt) {
483 char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN];
485 kprintf("redirect dst %s to %s\n",
486 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
487 dst_buf, INET_ADDRSTRLEN),
488 inet_ntop(AF_INET, &icp->icmp_gwaddr,
489 gw_buf, INET_ADDRSTRLEN));
492 icmpgw.sin_addr = ip->ip_src;
493 icmpdst.sin_addr = icp->icmp_gwaddr;
494 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
495 rtredirect_oncpu((struct sockaddr *)&icmpsrc,
496 (struct sockaddr *)&icmpdst, NULL, ICMP_RTREDIRECT_FLAGS,
497 (struct sockaddr *)&icmpgw);
498 kpfctlinput_direct(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
502 icmp_redirect_done_handler(netmsg_t nmsg)
504 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
505 struct mbuf *m = msg->m;
506 int hlen = msg->hlen;
508 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
509 struct icmp *icp = mtodoff(m, struct icmp *, hlen);;
511 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
512 key_sa_routechange((struct sockaddr *)&icmpsrc);
514 rip_input(&m, &hlen, msg->proto);
518 icmp_redirect_done(struct mbuf *m, int hlen, boolean_t dispatch_rip)
520 struct rt_addrinfo rtinfo;
521 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
522 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
523 struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
527 ip = mtod(m, struct ip *);
528 icp = mtodoff(m, struct icmp *, hlen);
530 icmpgw.sin_addr = ip->ip_src;
531 icmpdst.sin_addr = icp->icmp_gwaddr;
532 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
534 bzero(&rtinfo, sizeof(struct rt_addrinfo));
535 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&icmpsrc;
536 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&icmpdst;
537 rtinfo.rti_info[RTAX_NETMASK] = NULL;
538 rtinfo.rti_info[RTAX_AUTHOR] = (struct sockaddr *)&icmpgw;
539 rt_missmsg(RTM_REDIRECT, &rtinfo, ICMP_RTREDIRECT_FLAGS, 0);
542 struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg;
544 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
545 icmp_redirect_done_handler);
546 lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg);
551 icmp_redirect_handler(netmsg_t nmsg)
553 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
556 icmp_redirect(msg->m, msg->hlen, FALSE);
558 nextcpu = mycpuid + 1;
560 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg);
562 icmp_redirect_done(msg->m, msg->hlen, TRUE);
566 icmp_redirect_start(struct mbuf *m, int hlen, int proto)
568 struct netmsg_ctlinput *msg;
570 KASSERT(&curthread->td_msgport == netisr_cpuport(0),
573 icmp_redirect(m, hlen, TRUE);
576 /* There is only one netisr; done */
577 icmp_redirect_done(m, hlen, FALSE);
581 msg = &m->m_hdr.mh_ctlmsg;
582 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
583 icmp_redirect_handler);
585 msg->cmd = PRC_REDIRECT_HOST;
589 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
594 * Process a received ICMP message.
597 icmp_input(struct mbuf **mp, int *offp, int proto)
599 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
600 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
602 struct in_ifaddr *ia;
603 struct mbuf *m = *mp;
604 struct ip *ip = mtod(m, struct ip *);
605 int icmplen = ip->ip_len;
609 KASSERT(&curthread->td_msgport == netisr_cpuport(0),
616 * Locate icmp structure in mbuf, and check
617 * that not corrupted and of at least minimum length.
621 char src_buf[INET_ADDRSTRLEN], dst_buf[INET_ADDRSTRLEN];
623 kprintf("icmp_input from %s to %s, len %d\n",
624 inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN),
625 inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN),
629 if (icmplen < ICMP_MINLEN) {
630 icmpstat.icps_tooshort++;
633 i = hlen + min(icmplen, ICMP_ADVLENMIN);
634 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
635 icmpstat.icps_tooshort++;
636 return(IPPROTO_DONE);
638 ip = mtod(m, struct ip *);
640 if (in_cksum_skip(m, hlen + icmplen, hlen)) {
641 icmpstat.icps_checksum++;
644 icp = (struct icmp *)((caddr_t)ip + hlen);
646 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
648 * Deliver very specific ICMP type only.
650 switch (icp->icmp_type) {
661 kprintf("icmp_input, type %d code %d\n", icp->icmp_type,
666 * Message type specific processing.
668 if (icp->icmp_type > ICMP_MAXTYPE)
670 icmpstat.icps_inhist[icp->icmp_type]++;
671 code = icp->icmp_code;
672 switch (icp->icmp_type) {
676 case ICMP_UNREACH_NET:
677 case ICMP_UNREACH_HOST:
678 case ICMP_UNREACH_SRCFAIL:
679 case ICMP_UNREACH_NET_UNKNOWN:
680 case ICMP_UNREACH_HOST_UNKNOWN:
681 case ICMP_UNREACH_ISOLATED:
682 case ICMP_UNREACH_TOSNET:
683 case ICMP_UNREACH_TOSHOST:
684 case ICMP_UNREACH_HOST_PRECEDENCE:
685 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
686 code = PRC_UNREACH_NET;
689 case ICMP_UNREACH_NEEDFRAG:
694 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
695 * Treat subcodes 2,3 as immediate RST
697 case ICMP_UNREACH_PROTOCOL:
698 case ICMP_UNREACH_PORT:
699 code = PRC_UNREACH_PORT;
702 case ICMP_UNREACH_NET_PROHIB:
703 case ICMP_UNREACH_HOST_PROHIB:
704 case ICMP_UNREACH_FILTER_PROHIB:
705 code = PRC_UNREACH_ADMIN_PROHIB;
716 code += PRC_TIMXCEED_INTRANS;
722 code = PRC_PARAMPROB;
725 case ICMP_SOURCEQUENCH:
728 if (discard_sourcequench)
733 * Problem with datagram; advise higher level routines.
735 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
736 IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
737 icmpstat.icps_badlen++;
740 /* Discard ICMP's in response to multicast packets */
741 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
745 kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
747 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
752 if (code == PRC_MSGSIZE) {
753 /* Run MTU discovery in all netisrs */
754 if (icmp_mtudisc_start(m, hlen, proto)) {
755 /* Forwarded; done */
758 /* Move on; run rip_input() directly */
761 struct lwkt_port *port;
764 pr = &inetsw[ip_protox[icp->icmp_ip.ip_p]];
765 port = so_pr_ctlport(pr, code,
766 (struct sockaddr *)&icmpsrc, &icp->icmp_ip, &cpu);
771 * Run pr_ctlinput in all
774 icmp_ctlinput_global_start(m,
779 * There is only one netisr; run
780 * pr_ctlinput directly.
782 } else if (cpu != mycpuid) {
784 * Send to the target netisr to run
787 icmp_ctlinput_start(m, port,
793 * The target netisr is this netisr.
795 * XXX if the packet contains [IPv4 AH TCP],
796 * we can't make a notification to TCP layer.
798 so_pr_ctlinput_direct(pr, code,
799 (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
801 /* Move on; run rip_input() directly */
805 icmpstat.icps_badcode++;
810 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
811 icmpstat.icps_bmcastecho++;
814 icp->icmp_type = ICMP_ECHOREPLY;
816 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
824 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
825 icmpstat.icps_bmcasttstamp++;
828 if (icmplen < ICMP_TSLEN) {
829 icmpstat.icps_badlen++;
832 icp->icmp_type = ICMP_TSTAMPREPLY;
833 icp->icmp_rtime = iptime();
834 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
836 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
843 if (icmpmaskrepl == 0)
846 * We are not able to respond with all ones broadcast
847 * unless we receive it over a point-to-point interface.
849 if (icmplen < ICMP_MASKLEN)
851 switch (ip->ip_dst.s_addr) {
853 case INADDR_BROADCAST:
855 icmpdst.sin_addr = ip->ip_src;
859 icmpdst.sin_addr = ip->ip_dst;
861 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
862 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
867 icp->icmp_type = ICMP_MASKREPLY;
868 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
869 if (ip->ip_src.s_addr == 0) {
870 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
871 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
872 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
873 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
876 ip->ip_len += hlen; /* since ip_input deducts this */
877 icmpstat.icps_reflect++;
878 icmpstat.icps_outhist[icp->icmp_type]++;
880 return(IPPROTO_DONE);
884 char src_buf[INET_ADDRSTRLEN];
885 char dst_buf[INET_ADDRSTRLEN];
886 char gwy_buf[INET_ADDRSTRLEN];
888 kprintf("icmp redirect from %s: %s => %s\n",
889 inet_ntop(AF_INET, &ip->ip_src,
890 src_buf, INET_ADDRSTRLEN),
891 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
892 dst_buf, INET_ADDRSTRLEN),
893 inet_ntop(AF_INET, &icp->icmp_gwaddr,
894 gwy_buf, INET_ADDRSTRLEN));
900 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
901 IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
902 icmpstat.icps_badlen++;
907 char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN];
909 kprintf("redirect dst %s to %s\n",
910 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
911 dst_buf, INET_ADDRSTRLEN),
912 inet_ntop(AF_INET, &icp->icmp_gwaddr,
913 gw_buf, INET_ADDRSTRLEN));
916 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
918 /* Run redirect in all netisrs */
919 if (icmp_redirect_start(m, hlen, proto)) {
920 /* Forwarded; done */
923 /* Move on; run rip_input() directly */
925 key_sa_routechange((struct sockaddr *)&icmpsrc);
930 * No kernel processing for the following;
931 * just fall through to send to raw listener.
934 case ICMP_ROUTERADVERT:
935 case ICMP_ROUTERSOLICIT:
936 case ICMP_TSTAMPREPLY:
945 rip_input(mp, offp, proto);
946 return(IPPROTO_DONE);
950 return(IPPROTO_DONE);
954 * Reflect the ip packet back to the source
957 icmp_reflect(struct mbuf *m)
959 struct ip *ip = mtod(m, struct ip *);
960 struct in_ifaddr *ia;
961 struct in_ifaddr_container *iac;
962 struct ifaddr_container *ifac;
965 struct mbuf *opts = NULL;
966 int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
967 struct route *ro = NULL, rt;
969 if (!in_canforward(ip->ip_src) &&
970 ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
971 (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
972 m_freem(m); /* Bad return address */
973 icmpstat.icps_badaddr++;
974 goto done; /* Ip_output() will check for broadcast */
977 ip->ip_dst = ip->ip_src;
980 bzero(ro, sizeof *ro);
983 * If the incoming packet was addressed directly to us,
984 * use dst as the src for the reply. Otherwise (broadcast
985 * or anonymous), use the address which corresponds
986 * to the incoming interface.
989 LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
990 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
995 ifp = m->m_pkthdr.rcvif;
996 if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
997 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
998 struct ifaddr *ifa = ifac->ifa;
1000 if (ifa->ifa_addr->sa_family != AF_INET)
1003 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
1009 * If the packet was transiting through us, use the address of
1010 * the interface the packet came through in. If that interface
1011 * doesn't have a suitable IP address, the normal selection
1014 if (icmp_rfi && ifp != NULL) {
1015 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1016 struct ifaddr *ifa = ifac->ifa;
1018 if (ifa->ifa_addr->sa_family != AF_INET)
1025 * If the incoming packet was not addressed directly to us, use
1026 * designated interface for icmp replies specified by sysctl
1027 * net.inet.icmp.reply_src (default not set). Otherwise continue
1028 * with normal source selection.
1030 if (icmp_reply_src[0] != '\0' &&
1031 (ifp = ifunit_netisr(icmp_reply_src))) {
1032 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1033 struct ifaddr *ifa = ifac->ifa;
1035 if (ifa->ifa_addr->sa_family != AF_INET)
1042 * If the packet was transiting through us, use the address of
1043 * the interface that is the closest to the packet source.
1044 * When we don't have a route back to the packet source, stop here
1045 * and drop the packet.
1047 ia = ip_rtaddr(ip->ip_dst, ro);
1050 icmpstat.icps_noroute++;
1054 t = IA_SIN(ia)->sin_addr;
1056 ip->ip_ttl = ip_defttl;
1064 * Retrieve any source routing from the incoming packet;
1065 * add on any record-route or timestamp options.
1067 cp = (u_char *) (ip + 1);
1068 if ((opts = ip_srcroute(m)) == NULL &&
1069 (opts = m_gethdr(M_NOWAIT, MT_HEADER))) {
1070 opts->m_len = sizeof(struct in_addr);
1071 mtod(opts, struct in_addr *)->s_addr = 0;
1076 kprintf("icmp_reflect optlen %d rt %d => ",
1077 optlen, opts->m_len);
1079 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
1080 opt = cp[IPOPT_OPTVAL];
1081 if (opt == IPOPT_EOL)
1083 if (opt == IPOPT_NOP)
1086 if (cnt < IPOPT_OLEN + sizeof *cp)
1088 len = cp[IPOPT_OLEN];
1089 if (len < IPOPT_OLEN + sizeof *cp ||
1094 * Should check for overflow, but it
1097 if (opt == IPOPT_RR || opt == IPOPT_TS ||
1098 opt == IPOPT_SECURITY) {
1100 mtod(opts, caddr_t) + opts->m_len,
1105 /* Terminate & pad, if necessary */
1106 cnt = opts->m_len % 4;
1108 for (; cnt < 4; cnt++) {
1109 *(mtod(opts, caddr_t) + opts->m_len) =
1116 kprintf("%d\n", opts->m_len);
1120 * Now strip out original options by copying rest of first
1121 * mbuf's data back, and adjust the IP length.
1123 ip->ip_len -= optlen;
1124 ip->ip_vhl = IP_VHL_BORING;
1126 if (m->m_flags & M_PKTHDR)
1127 m->m_pkthdr.len -= optlen;
1128 optlen += sizeof(struct ip);
1129 bcopy((caddr_t)ip + optlen, ip + 1,
1130 m->m_len - sizeof(struct ip));
1132 m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
1133 m->m_flags &= ~(M_BCAST|M_MCAST);
1134 icmp_send(m, opts, ro);
1138 if (ro && ro->ro_rt)
1143 * Send an icmp packet back to the ip level,
1144 * after supplying a checksum.
1147 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
1149 struct ip *ip = mtod(m, struct ip *);
1153 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1156 icp = mtod(m, struct icmp *);
1157 icp->icmp_cksum = 0;
1158 icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
1161 m->m_pkthdr.rcvif = NULL;
1164 char dst_buf[INET_ADDRSTRLEN], src_buf[INET_ADDRSTRLEN];
1166 kprintf("icmp_send dst %s src %s\n",
1167 inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN),
1168 inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN));
1171 ip_output(m, opts, rt, 0, NULL, NULL);
1181 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1187 * Return the next larger or smaller MTU plateau (table from RFC 1191)
1188 * given current value MTU. If DIR is less than zero, a larger plateau
1189 * is returned; otherwise, a smaller value is returned.
1192 ip_next_mtu(int mtu, int dir)
1194 static int mtutab[] = {
1195 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
1200 for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
1201 if (mtu >= mtutab[i])
1209 return mtutab[i - 1];
1212 if (mtutab[i] == 0) {
1214 } else if(mtu > mtutab[i]) {
1217 return mtutab[i + 1];
1225 * badport_bandlim() - check for ICMP bandwidth limit
1227 * Return 0 if it is ok to send an ICMP error response, -1 if we have
1228 * hit our bandwidth limit and it is not ok.
1230 * If icmplim is <= 0, the feature is disabled and 0 is returned.
1232 * For now we separate the TCP and UDP subsystems w/ different 'which'
1233 * values. We may eventually remove this separation (and simplify the
1236 * Note that the printing of the error message is delayed so we can
1237 * properly print the icmp error rate that the system was trying to do
1238 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing
1239 * the 'final' error, but it doesn't make sense to solve the printing
1240 * delay with more complex code.
1243 badport_bandlim(int which)
1245 static int lticks[BANDLIM_MAX + 1];
1246 static int lpackets[BANDLIM_MAX + 1];
1248 const char *bandlimittype[] = {
1249 "Limiting icmp unreach response",
1250 "Limiting icmp ping response",
1251 "Limiting icmp tstamp response",
1252 "Limiting closed port RST response",
1253 "Limiting open port RST response"
1257 * Return ok status if feature disabled or argument out of
1261 if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
1263 dticks = ticks - lticks[which];
1266 * reset stats when cumulative dt exceeds one second.
1269 if ((unsigned int)dticks > hz) {
1270 if (lpackets[which] > icmplim && icmplim_output) {
1271 kprintf("%s from %d to %d packets per second\n",
1272 bandlimittype[which],
1277 lticks[which] = ticks;
1278 lpackets[which] = 0;
1285 if (++lpackets[which] > icmplim) {