3b3b878649e3290c04fcbb30d57b8bf73e82d357
[dragonfly.git] / sys / netinet / ip_icmp.c
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *      The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
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.
16  *
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
27  * SUCH DAMAGE.
28  *
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 $
31  */
32
33 #include "opt_ipsec.h"
34
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/mbuf.h>
38 #include <sys/protosw.h>
39 #include <sys/socket.h>
40 #include <sys/socketops.h>
41 #include <sys/time.h>
42 #include <sys/kernel.h>
43 #include <sys/sysctl.h>
44 #include <sys/in_cksum.h>
45
46 #include <machine/stdarg.h>
47
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/route.h>
51
52 #define _IP_VHL
53 #include <netinet/in.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/icmp_var.h>
60
61 #ifdef IPSEC
62 #include <netinet6/ipsec.h>
63 #include <netproto/key/key.h>
64 #endif
65
66 #ifdef FAST_IPSEC
67 #include <netproto/ipsec/ipsec.h>
68 #include <netproto/ipsec/key.h>
69 #define IPSEC
70 #endif
71
72 /*
73  * ICMP routines: error generation, receive packet processing, and
74  * routines to turnaround packets back to the originator, and
75  * host table maintenance routines.
76  */
77
78 struct icmpstat icmpstat;
79 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
80         &icmpstat, icmpstat, "ICMP statistics");
81
82 static int      icmpmaskrepl = 0;
83 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
84         &icmpmaskrepl, 0, "Allow replies to netmask requests");
85
86 static int      drop_redirect = 0;
87 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
88         &drop_redirect, 0, "Ignore ICMP redirects");
89
90 static int      log_redirect = 0;
91 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
92         &log_redirect, 0, "Enable output about ICMP redirects");
93
94 static int      discard_sourcequench = 1;
95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, discard_sourcequench, CTLFLAG_RW,
96         &discard_sourcequench, 0, "Discard ICMP Source Quench");
97
98 #ifdef ICMP_BANDLIM
99
100 /*
101  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
102  *      variable content is -1 and read-only.
103  */
104
105 static int      icmplim = 200;
106 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
107         &icmplim, 0, "ICMP bandwidth limit");
108 #else
109
110 static int      icmplim = -1;
111 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
112         &icmplim, 0, "ICMP bandwidth limit");
113         
114 #endif
115
116 static int      icmplim_output = 0;
117 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
118         &icmplim_output, 0, "Enable output about ICMP bandwidth limits");
119
120 /*
121  * ICMP broadcast echo sysctl
122  */
123
124 static int      icmpbmcastecho = 0;
125 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
126     &icmpbmcastecho, 0, "");
127
128 static char     icmp_reply_src[IFNAMSIZ];
129 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
130         icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
131
132 static int      icmp_rfi;
133 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
134         &icmp_rfi, 0, "ICMP reply from incoming interface for "
135         "non-local packets");
136
137 #ifdef ICMPPRINTFS
138 int     icmpprintfs = 0;
139 #endif
140
141 static void     icmp_reflect (struct mbuf *);
142 static void     icmp_send (struct mbuf *, struct mbuf *, struct route *);
143
144 extern  struct protosw inetsw[];
145
146 /*
147  * Generate an error packet of type error
148  * in response to bad packet ip.
149  */
150 void
151 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
152 {
153         struct ip *oip = mtod(n, struct ip *), *nip;
154         unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
155         struct icmp *icp;
156         struct mbuf *m;
157         unsigned icmplen;
158
159 #ifdef ICMPPRINTFS
160         if (icmpprintfs)
161                 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
162 #endif
163         if (type != ICMP_REDIRECT)
164                 icmpstat.icps_error++;
165         /*
166          * Don't send error if the original packet was encrypted.
167          * Don't send error if not the first fragment of message.
168          * Don't error if the old packet protocol was ICMP
169          * error message, only known informational types.
170          */
171         if (n->m_flags & M_DECRYPTED)
172                 goto freeit;
173         if (oip->ip_off &~ (IP_MF|IP_DF))
174                 goto freeit;
175         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
176           n->m_len >= oiplen + ICMP_MINLEN &&
177           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
178                 icmpstat.icps_oldicmp++;
179                 goto freeit;
180         }
181         /* Don't send error in response to a multicast or broadcast packet */
182         if (n->m_flags & (M_BCAST|M_MCAST))
183                 goto freeit;
184         /*
185          * First, formulate icmp message
186          */
187         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
188         if (m == NULL)
189                 goto freeit;
190         icmplen = min(oiplen + 8, oip->ip_len);
191         if (icmplen < sizeof(struct ip))
192                 panic("icmp_error: bad length");
193         m->m_len = icmplen + ICMP_MINLEN;
194         MH_ALIGN(m, m->m_len);
195         icp = mtod(m, struct icmp *);
196         if ((u_int)type > ICMP_MAXTYPE)
197                 panic("icmp_error");
198         icmpstat.icps_outhist[type]++;
199         icp->icmp_type = type;
200         if (type == ICMP_REDIRECT)
201                 icp->icmp_gwaddr.s_addr = dest;
202         else {
203                 icp->icmp_void = 0;
204                 /*
205                  * The following assignments assume an overlay with the
206                  * zeroed icmp_void field.
207                  */
208                 if (type == ICMP_PARAMPROB) {
209                         icp->icmp_pptr = code;
210                         code = 0;
211                 } else if (type == ICMP_UNREACH &&
212                         code == ICMP_UNREACH_NEEDFRAG && destmtu) {
213                         icp->icmp_nextmtu = htons(destmtu);
214                 }
215         }
216
217         icp->icmp_code = code;
218         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
219         nip = &icp->icmp_ip;
220
221         /*
222          * Convert fields to network representation.
223          */
224         nip->ip_len = htons(nip->ip_len);
225         nip->ip_off = htons(nip->ip_off);
226
227         /*
228          * Now, copy old ip header (without options)
229          * in front of icmp message.
230          */
231         if (m->m_data - sizeof(struct ip) < m->m_pktdat)
232                 panic("icmp len");
233         m->m_data -= sizeof(struct ip);
234         m->m_len += sizeof(struct ip);
235         m->m_pkthdr.len = m->m_len;
236         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
237         nip = mtod(m, struct ip *);
238         bcopy(oip, nip, sizeof(struct ip));
239         nip->ip_len = m->m_len;
240         nip->ip_vhl = IP_VHL_BORING;
241         nip->ip_p = IPPROTO_ICMP;
242         nip->ip_tos = 0;
243         m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
244         icmp_reflect(m);
245
246 freeit:
247         m_freem(n);
248 }
249
250 /*
251  * Process a received ICMP message.
252  */
253 int
254 icmp_input(struct mbuf **mp, int *offp, int proto)
255 {
256         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
257         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
258         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
259         struct icmp *icp;
260         struct in_ifaddr *ia;
261         struct mbuf *m = *mp;
262         struct ip *ip = mtod(m, struct ip *);
263         int icmplen = ip->ip_len;
264         int i, hlen;
265         int code;
266
267         *mp = NULL;
268         hlen = *offp;
269
270         /*
271          * Locate icmp structure in mbuf, and check
272          * that not corrupted and of at least minimum length.
273          */
274 #ifdef ICMPPRINTFS
275         if (icmpprintfs) {
276                 char buf[sizeof "aaa.bbb.ccc.ddd"];
277
278                 strcpy(buf, inet_ntoa(ip->ip_src));
279                 kprintf("icmp_input from %s to %s, len %d\n",
280                        buf, inet_ntoa(ip->ip_dst), icmplen);
281         }
282 #endif
283         if (icmplen < ICMP_MINLEN) {
284                 icmpstat.icps_tooshort++;
285                 goto freeit;
286         }
287         i = hlen + min(icmplen, ICMP_ADVLENMIN);
288         if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
289                 icmpstat.icps_tooshort++;
290                 return(IPPROTO_DONE);
291         }
292         ip = mtod(m, struct ip *);
293
294         if (in_cksum_skip(m, hlen + icmplen, hlen)) {
295                 icmpstat.icps_checksum++;
296                 goto freeit;
297         }
298         icp = mtod(m, struct icmp *);
299
300         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
301                 /*
302                  * Deliver very specific ICMP type only.
303                  */
304                 switch (icp->icmp_type) {
305                 case ICMP_UNREACH:
306                 case ICMP_TIMXCEED:
307                         break;
308                 default:
309                         goto freeit;
310                 }
311         }
312
313 #ifdef ICMPPRINTFS
314         if (icmpprintfs)
315                 kprintf("icmp_input, type %d code %d\n", icp->icmp_type,
316                     icp->icmp_code);
317 #endif
318
319         /*
320          * Message type specific processing.
321          */
322         if (icp->icmp_type > ICMP_MAXTYPE)
323                 goto raw;
324         icmpstat.icps_inhist[icp->icmp_type]++;
325         code = icp->icmp_code;
326         switch (icp->icmp_type) {
327
328         case ICMP_UNREACH:
329                 switch (code) {
330                         case ICMP_UNREACH_NET:
331                         case ICMP_UNREACH_HOST:
332                         case ICMP_UNREACH_SRCFAIL:
333                         case ICMP_UNREACH_NET_UNKNOWN:
334                         case ICMP_UNREACH_HOST_UNKNOWN:
335                         case ICMP_UNREACH_ISOLATED:
336                         case ICMP_UNREACH_TOSNET:
337                         case ICMP_UNREACH_TOSHOST:
338                         case ICMP_UNREACH_HOST_PRECEDENCE:
339                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
340                                 code = PRC_UNREACH_NET;
341                                 break;
342
343                         case ICMP_UNREACH_NEEDFRAG:
344                                 code = PRC_MSGSIZE;
345                                 break;
346
347                         /*
348                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
349                          * Treat subcodes 2,3 as immediate RST
350                          */
351                         case ICMP_UNREACH_PROTOCOL:
352                         case ICMP_UNREACH_PORT:
353                                 code = PRC_UNREACH_PORT;
354                                 break;
355
356                         case ICMP_UNREACH_NET_PROHIB:
357                         case ICMP_UNREACH_HOST_PROHIB:
358                         case ICMP_UNREACH_FILTER_PROHIB:
359                                 code = PRC_UNREACH_ADMIN_PROHIB;
360                                 break;
361
362                         default:
363                                 goto badcode;
364                 }
365                 goto deliver;
366
367         case ICMP_TIMXCEED:
368                 if (code > 1)
369                         goto badcode;
370                 code += PRC_TIMXCEED_INTRANS;
371                 goto deliver;
372
373         case ICMP_PARAMPROB:
374                 if (code > 1)
375                         goto badcode;
376                 code = PRC_PARAMPROB;
377                 goto deliver;
378
379         case ICMP_SOURCEQUENCH:
380                 if (code)
381                         goto badcode;
382                 if (discard_sourcequench)
383                         break;
384                 code = PRC_QUENCH;
385 deliver:
386                 /*
387                  * Problem with datagram; advise higher level routines.
388                  */
389                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
390                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
391                         icmpstat.icps_badlen++;
392                         goto freeit;
393                 }
394                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
395                 /* Discard ICMP's in response to multicast packets */
396                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
397                         goto badcode;
398 #ifdef ICMPPRINTFS
399                 if (icmpprintfs)
400                         kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
401 #endif
402                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
403 #if 1
404                 /*
405                  * MTU discovery:
406                  * If we got a needfrag and there is a host route to the
407                  * original destination, and the MTU is not locked, then
408                  * set the MTU in the route to the suggested new value
409                  * (if given) and then notify as usual.  The ULPs will
410                  * notice that the MTU has changed and adapt accordingly.
411                  * If no new MTU was suggested, then we guess a new one
412                  * less than the current value.  If the new MTU is
413                  * unreasonably small (arbitrarily set at 296), then
414                  * we reset the MTU to the interface value and enable the
415                  * lock bit, indicating that we are no longer doing MTU
416                  * discovery.
417                  */
418                 if (code == PRC_MSGSIZE) {
419                         struct rtentry *rt;
420                         int mtu;
421
422                         rt = rtpurelookup((struct sockaddr *)&icmpsrc);
423                         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
424                             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
425                                 mtu = ntohs(icp->icmp_nextmtu);
426                                 if (!mtu)
427                                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
428                                                           1);
429 #ifdef DEBUG_MTUDISC
430                                 kprintf("MTU for %s reduced to %d\n",
431                                         inet_ntoa(icmpsrc.sin_addr), mtu);
432 #endif
433                                 if (mtu < 296) {
434                                         /* rt->rt_rmx.rmx_mtu =
435                                                 rt->rt_ifp->if_mtu; */
436                                         rt->rt_rmx.rmx_locks |= RTV_MTU;
437                                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
438                                         rt->rt_rmx.rmx_mtu = mtu;
439                                 }
440                         }
441                         if (rt != NULL)
442                                 --rt->rt_refcnt;
443                 }
444 #endif
445                 /*
446                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
447                  * notification to TCP layer.
448                  */
449                 so_pru_ctlinput(
450                         &inetsw[ip_protox[icp->icmp_ip.ip_p]],
451                         code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
452                 break;
453
454 badcode:
455                 icmpstat.icps_badcode++;
456                 break;
457
458         case ICMP_ECHO:
459                 if (!icmpbmcastecho
460                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
461                         icmpstat.icps_bmcastecho++;
462                         break;
463                 }
464                 icp->icmp_type = ICMP_ECHOREPLY;
465 #ifdef ICMP_BANDLIM
466                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
467                         goto freeit;
468                 else
469 #endif
470                         goto reflect;
471
472         case ICMP_TSTAMP:
473                 if (!icmpbmcastecho
474                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
475                         icmpstat.icps_bmcasttstamp++;
476                         break;
477                 }
478                 if (icmplen < ICMP_TSLEN) {
479                         icmpstat.icps_badlen++;
480                         break;
481                 }
482                 icp->icmp_type = ICMP_TSTAMPREPLY;
483                 icp->icmp_rtime = iptime();
484                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
485 #ifdef ICMP_BANDLIM
486                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
487                         goto freeit;
488                 else
489 #endif
490                         goto reflect;
491
492         case ICMP_MASKREQ:
493                 if (icmpmaskrepl == 0)
494                         break;
495                 /*
496                  * We are not able to respond with all ones broadcast
497                  * unless we receive it over a point-to-point interface.
498                  */
499                 if (icmplen < ICMP_MASKLEN)
500                         break;
501                 switch (ip->ip_dst.s_addr) {
502
503                 case INADDR_BROADCAST:
504                 case INADDR_ANY:
505                         icmpdst.sin_addr = ip->ip_src;
506                         break;
507
508                 default:
509                         icmpdst.sin_addr = ip->ip_dst;
510                 }
511                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
512                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
513                 if (ia == NULL)
514                         break;
515                 if (ia->ia_ifp == 0)
516                         break;
517                 icp->icmp_type = ICMP_MASKREPLY;
518                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
519                 if (ip->ip_src.s_addr == 0) {
520                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
521                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
522                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
523                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
524                 }
525 reflect:
526                 ip->ip_len += hlen;     /* since ip_input deducts this */
527                 icmpstat.icps_reflect++;
528                 icmpstat.icps_outhist[icp->icmp_type]++;
529                 icmp_reflect(m);
530                 return(IPPROTO_DONE);
531
532         case ICMP_REDIRECT:
533                 if (log_redirect) {
534                         u_long src, dst, gw;
535
536                         src = ntohl(ip->ip_src.s_addr);
537                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
538                         gw = ntohl(icp->icmp_gwaddr.s_addr);
539                         kprintf("icmp redirect from %d.%d.%d.%d: "
540                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
541                                (int)(src >> 24), (int)((src >> 16) & 0xff),
542                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
543                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
544                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
545                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
546                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
547                 }
548                 if (drop_redirect)
549                         break;
550                 if (code > 3)
551                         goto badcode;
552                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
553                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
554                         icmpstat.icps_badlen++;
555                         break;
556                 }
557                 /*
558                  * Short circuit routing redirects to force
559                  * immediate change in the kernel's routing
560                  * tables.  The message is also handed to anyone
561                  * listening on a raw socket (e.g. the routing
562                  * daemon for use in updating its tables).
563                  */
564                 icmpgw.sin_addr = ip->ip_src;
565                 icmpdst.sin_addr = icp->icmp_gwaddr;
566 #ifdef  ICMPPRINTFS
567                 if (icmpprintfs) {
568                         char buf[sizeof "aaa.bbb.ccc.ddd"];
569
570                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
571                         kprintf("redirect dst %s to %s\n",
572                                buf, inet_ntoa(icp->icmp_gwaddr));
573                 }
574 #endif
575                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
576                 rtredirect((struct sockaddr *)&icmpsrc,
577                   (struct sockaddr *)&icmpdst,
578                   NULL, RTF_GATEWAY | RTF_HOST,
579                   (struct sockaddr *)&icmpgw);
580                 kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
581 #ifdef IPSEC
582                 key_sa_routechange((struct sockaddr *)&icmpsrc);
583 #endif
584                 break;
585
586         /*
587          * No kernel processing for the following;
588          * just fall through to send to raw listener.
589          */
590         case ICMP_ECHOREPLY:
591         case ICMP_ROUTERADVERT:
592         case ICMP_ROUTERSOLICIT:
593         case ICMP_TSTAMPREPLY:
594         case ICMP_IREQREPLY:
595         case ICMP_MASKREPLY:
596         default:
597                 break;
598         }
599
600 raw:
601         *mp = m;
602         rip_input(mp, offp, proto);
603         return(IPPROTO_DONE);
604
605 freeit:
606         m_freem(m);
607         return(IPPROTO_DONE);
608 }
609
610 /*
611  * Reflect the ip packet back to the source
612  */
613 static void
614 icmp_reflect(struct mbuf *m)
615 {
616         struct ip *ip = mtod(m, struct ip *);
617         struct in_ifaddr *ia;
618         struct in_ifaddr_container *iac;
619         struct ifaddr_container *ifac;
620         struct ifnet *ifp;
621         struct in_addr t;
622         struct mbuf *opts = NULL;
623         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
624         struct route *ro = NULL, rt;
625
626         if (!in_canforward(ip->ip_src) &&
627             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
628              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
629                 m_freem(m);     /* Bad return address */
630                 icmpstat.icps_badaddr++;
631                 goto done;      /* Ip_output() will check for broadcast */
632         }
633         t = ip->ip_dst;
634         ip->ip_dst = ip->ip_src;
635
636         ro = &rt;
637         bzero(ro, sizeof *ro);
638
639         /*
640          * If the incoming packet was addressed directly to us,
641          * use dst as the src for the reply.  Otherwise (broadcast
642          * or anonymous), use the address which corresponds
643          * to the incoming interface.
644          */
645         ia = NULL;
646         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
647                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
648                         ia = iac->ia;
649                         goto match;
650                 }
651         }
652         ifp = m->m_pkthdr.rcvif;
653         if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
654                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
655                         struct ifaddr *ifa = ifac->ifa;
656
657                         if (ifa->ifa_addr->sa_family != AF_INET)
658                                 continue;
659                         ia = ifatoia(ifa);
660                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
661                             t.s_addr)
662                                 goto match;
663                 }
664         }
665         /*
666          * If the packet was transiting through us, use the address of
667          * the interface the packet came through in.  If that interface
668          * doesn't have a suitable IP address, the normal selection
669          * criteria apply.
670          */
671         if (icmp_rfi && ifp != NULL) {
672                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
673                         struct ifaddr *ifa = ifac->ifa;
674
675                         if (ifa->ifa_addr->sa_family != AF_INET)
676                                 continue;
677                         ia = ifatoia(ifa);
678                         goto match;
679                 }
680         }
681         /*
682          * If the incoming packet was not addressed directly to us, use
683          * designated interface for icmp replies specified by sysctl
684          * net.inet.icmp.reply_src (default not set). Otherwise continue
685          * with normal source selection.
686          */
687         if (icmp_reply_src[0] != '\0' && (ifp = ifunit(icmp_reply_src))) {
688                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
689                         struct ifaddr *ifa = ifac->ifa;
690
691                         if (ifa->ifa_addr->sa_family != AF_INET)
692                                 continue;
693                         ia = ifatoia(ifa);
694                         goto match;
695                 }
696         }
697         /*
698          * If the packet was transiting through us, use the address of
699          * the interface that is the closest to the packet source.
700          * When we don't have a route back to the packet source, stop here
701          * and drop the packet.
702          */
703         ia = ip_rtaddr(ip->ip_dst, ro);
704         if (ia == NULL) {
705                 m_freem(m);
706                 icmpstat.icps_noroute++;
707                 goto done;
708         }
709 match:
710         t = IA_SIN(ia)->sin_addr;
711         ip->ip_src = t;
712         ip->ip_ttl = ip_defttl;
713
714         if (optlen > 0) {
715                 u_char *cp;
716                 int opt, cnt;
717                 u_int len;
718
719                 /*
720                  * Retrieve any source routing from the incoming packet;
721                  * add on any record-route or timestamp options.
722                  */
723                 cp = (u_char *) (ip + 1);
724                 if ((opts = ip_srcroute(m)) == NULL &&
725                     (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
726                         opts->m_len = sizeof(struct in_addr);
727                         mtod(opts, struct in_addr *)->s_addr = 0;
728                 }
729                 if (opts) {
730 #ifdef ICMPPRINTFS
731                         if (icmpprintfs)
732                                 kprintf("icmp_reflect optlen %d rt %d => ",
733                                        optlen, opts->m_len);
734 #endif
735                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
736                                 opt = cp[IPOPT_OPTVAL];
737                                 if (opt == IPOPT_EOL)
738                                         break;
739                                 if (opt == IPOPT_NOP)
740                                         len = 1;
741                                 else {
742                                         if (cnt < IPOPT_OLEN + sizeof *cp)
743                                                 break;
744                                         len = cp[IPOPT_OLEN];
745                                         if (len < IPOPT_OLEN + sizeof *cp ||
746                                             len > cnt)
747                                         break;
748                                 }
749                                 /*
750                                  * Should check for overflow, but it
751                                  * "can't happen".
752                                  */
753                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
754                                     opt == IPOPT_SECURITY) {
755                                         bcopy(cp,
756                                               mtod(opts, caddr_t) + opts->m_len,
757                                               len);
758                                         opts->m_len += len;
759                                 }
760                         }
761                         /* Terminate & pad, if necessary */
762                         cnt = opts->m_len % 4;
763                         if (cnt) {
764                                 for (; cnt < 4; cnt++) {
765                                         *(mtod(opts, caddr_t) + opts->m_len) =
766                                             IPOPT_EOL;
767                                         opts->m_len++;
768                                 }
769                         }
770 #ifdef ICMPPRINTFS
771                         if (icmpprintfs)
772                                 kprintf("%d\n", opts->m_len);
773 #endif
774                 }
775                 /*
776                  * Now strip out original options by copying rest of first
777                  * mbuf's data back, and adjust the IP length.
778                  */
779                 ip->ip_len -= optlen;
780                 ip->ip_vhl = IP_VHL_BORING;
781                 m->m_len -= optlen;
782                 if (m->m_flags & M_PKTHDR)
783                         m->m_pkthdr.len -= optlen;
784                 optlen += sizeof(struct ip);
785                 bcopy((caddr_t)ip + optlen, ip + 1,
786                       m->m_len - sizeof(struct ip));
787         }
788         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
789         m->m_flags &= ~(M_BCAST|M_MCAST);
790         icmp_send(m, opts, ro);
791 done:
792         if (opts)
793                 m_free(opts);
794         if (ro && ro->ro_rt)
795                 RTFREE(ro->ro_rt);
796 }
797
798 /*
799  * Send an icmp packet back to the ip level,
800  * after supplying a checksum.
801  */
802 static void
803 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
804 {
805         struct ip *ip = mtod(m, struct ip *);
806         struct icmp *icp;
807         int hlen;
808
809         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
810         m->m_data += hlen;
811         m->m_len -= hlen;
812         icp = mtod(m, struct icmp *);
813         icp->icmp_cksum = 0;
814         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
815         m->m_data -= hlen;
816         m->m_len += hlen;
817         m->m_pkthdr.rcvif = NULL;
818 #ifdef ICMPPRINTFS
819         if (icmpprintfs) {
820                 char buf[sizeof "aaa.bbb.ccc.ddd"];
821
822                 strcpy(buf, inet_ntoa(ip->ip_dst));
823                 kprintf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src));
824         }
825 #endif
826         ip_output(m, opts, rt, 0, NULL, NULL);
827 }
828
829 n_time
830 iptime(void)
831 {
832         struct timeval atv;
833         u_long t;
834
835         getmicrotime(&atv);
836         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
837         return (htonl(t));
838 }
839
840 #if 1
841 /*
842  * Return the next larger or smaller MTU plateau (table from RFC 1191)
843  * given current value MTU.  If DIR is less than zero, a larger plateau
844  * is returned; otherwise, a smaller value is returned.
845  */
846 int
847 ip_next_mtu(int mtu, int dir)
848 {
849         static int mtutab[] = {
850                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
851                 68, 0
852         };
853         int i;
854
855         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
856                 if (mtu >= mtutab[i])
857                         break;
858         }
859
860         if (dir < 0) {
861                 if (i == 0) {
862                         return 0;
863                 } else {
864                         return mtutab[i - 1];
865                 }
866         } else {
867                 if (mtutab[i] == 0) {
868                         return 0;
869                 } else if(mtu > mtutab[i]) {
870                         return mtutab[i];
871                 } else {
872                         return mtutab[i + 1];
873                 }
874         }
875 }
876 #endif
877
878 #ifdef ICMP_BANDLIM
879 /*
880  * badport_bandlim() - check for ICMP bandwidth limit
881  *
882  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
883  *      hit our bandwidth limit and it is not ok.
884  *
885  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
886  *
887  *      For now we separate the TCP and UDP subsystems w/ different 'which'
888  *      values.  We may eventually remove this separation (and simplify the
889  *      code further).
890  *
891  *      Note that the printing of the error message is delayed so we can
892  *      properly print the icmp error rate that the system was trying to do
893  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
894  *      the 'final' error, but it doesn't make sense to solve the printing
895  *      delay with more complex code.
896  */
897 int
898 badport_bandlim(int which)
899 {
900         static int lticks[BANDLIM_MAX + 1];
901         static int lpackets[BANDLIM_MAX + 1];
902         int dticks;
903         const char *bandlimittype[] = {
904                 "Limiting icmp unreach response",
905                 "Limiting icmp ping response",
906                 "Limiting icmp tstamp response",
907                 "Limiting closed port RST response",
908                 "Limiting open port RST response"
909                 };
910
911         /*
912          * Return ok status if feature disabled or argument out of
913          * ranage.
914          */
915
916         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
917                 return(0);
918         dticks = ticks - lticks[which];
919
920         /*
921          * reset stats when cumulative dt exceeds one second.
922          */
923
924         if ((unsigned int)dticks > hz) {
925                 if (lpackets[which] > icmplim && icmplim_output) {
926                         kprintf("%s from %d to %d packets per second\n",
927                                 bandlimittype[which],
928                                 lpackets[which],
929                                 icmplim
930                         );
931                 }
932                 lticks[which] = ticks;
933                 lpackets[which] = 0;
934         }
935
936         /*
937          * bump packet count
938          */
939
940         if (++lpackets[which] > icmplim) {
941                 return(-1);
942         }
943         return(0);
944 }
945 #endif