IA_PRF_RTEXISTOK is no longer needed, in_{add,scrub}prefix() does more
[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. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *      This product includes software developed by the University of
16  *      California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
34  * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
35  * $DragonFly: src/sys/netinet/ip_icmp.c,v 1.32 2008/10/27 02:56:30 sephe Exp $
36  */
37
38 #include "opt_ipsec.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/mbuf.h>
43 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/socketops.h>
46 #include <sys/time.h>
47 #include <sys/kernel.h>
48 #include <sys/sysctl.h>
49 #include <sys/in_cksum.h>
50
51 #include <machine/stdarg.h>
52
53 #include <net/if.h>
54 #include <net/if_types.h>
55 #include <net/route.h>
56
57 #define _IP_VHL
58 #include <netinet/in.h>
59 #include <netinet/in_systm.h>
60 #include <netinet/in_var.h>
61 #include <netinet/ip.h>
62 #include <netinet/ip_icmp.h>
63 #include <netinet/ip_var.h>
64 #include <netinet/icmp_var.h>
65
66 #ifdef IPSEC
67 #include <netinet6/ipsec.h>
68 #include <netproto/key/key.h>
69 #endif
70
71 #ifdef FAST_IPSEC
72 #include <netproto/ipsec/ipsec.h>
73 #include <netproto/ipsec/key.h>
74 #define IPSEC
75 #endif
76
77 /*
78  * ICMP routines: error generation, receive packet processing, and
79  * routines to turnaround packets back to the originator, and
80  * host table maintenance routines.
81  */
82
83 struct icmpstat icmpstat;
84 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
85         &icmpstat, icmpstat, "");
86
87 static int      icmpmaskrepl = 0;
88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
89         &icmpmaskrepl, 0, "");
90
91 static int      drop_redirect = 0;
92 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
93         &drop_redirect, 0, "");
94
95 static int      log_redirect = 0;
96 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
97         &log_redirect, 0, "");
98
99 #ifdef ICMP_BANDLIM
100
101 /*
102  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
103  *      variable content is -1 and read-only.
104  */
105
106 static int      icmplim = 200;
107 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
108         &icmplim, 0, "");
109 #else
110
111 static int      icmplim = -1;
112 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
113         &icmplim, 0, "");
114         
115 #endif
116
117 static int      icmplim_output = 1;
118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
119         &icmplim_output, 0, "");
120
121 /*
122  * ICMP broadcast echo sysctl
123  */
124
125 static int      icmpbmcastecho = 0;
126 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
127         &icmpbmcastecho, 0, "");
128
129
130 #ifdef ICMPPRINTFS
131 int     icmpprintfs = 0;
132 #endif
133
134 static void     icmp_reflect (struct mbuf *);
135 static void     icmp_send (struct mbuf *, struct mbuf *, struct route *);
136
137 extern  struct protosw inetsw[];
138
139 /*
140  * Generate an error packet of type error
141  * in response to bad packet ip.
142  */
143 void
144 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
145 {
146         struct ip *oip = mtod(n, struct ip *), *nip;
147         unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
148         struct icmp *icp;
149         struct mbuf *m;
150         unsigned icmplen;
151
152 #ifdef ICMPPRINTFS
153         if (icmpprintfs)
154                 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
155 #endif
156         if (type != ICMP_REDIRECT)
157                 icmpstat.icps_error++;
158         /*
159          * Don't send error if the original packet was encrypted.
160          * Don't send error if not the first fragment of message.
161          * Don't error if the old packet protocol was ICMP
162          * error message, only known informational types.
163          */
164         if (n->m_flags & M_DECRYPTED)
165                 goto freeit;
166         if (oip->ip_off &~ (IP_MF|IP_DF))
167                 goto freeit;
168         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
169           n->m_len >= oiplen + ICMP_MINLEN &&
170           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
171                 icmpstat.icps_oldicmp++;
172                 goto freeit;
173         }
174         /* Don't send error in response to a multicast or broadcast packet */
175         if (n->m_flags & (M_BCAST|M_MCAST))
176                 goto freeit;
177         /*
178          * First, formulate icmp message
179          */
180         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
181         if (m == NULL)
182                 goto freeit;
183         icmplen = min(oiplen + 8, oip->ip_len);
184         if (icmplen < sizeof(struct ip))
185                 panic("icmp_error: bad length");
186         m->m_len = icmplen + ICMP_MINLEN;
187         MH_ALIGN(m, m->m_len);
188         icp = mtod(m, struct icmp *);
189         if ((u_int)type > ICMP_MAXTYPE)
190                 panic("icmp_error");
191         icmpstat.icps_outhist[type]++;
192         icp->icmp_type = type;
193         if (type == ICMP_REDIRECT)
194                 icp->icmp_gwaddr.s_addr = dest;
195         else {
196                 icp->icmp_void = 0;
197                 /*
198                  * The following assignments assume an overlay with the
199                  * zeroed icmp_void field.
200                  */
201                 if (type == ICMP_PARAMPROB) {
202                         icp->icmp_pptr = code;
203                         code = 0;
204                 } else if (type == ICMP_UNREACH &&
205                         code == ICMP_UNREACH_NEEDFRAG && destmtu) {
206                         icp->icmp_nextmtu = htons(destmtu);
207                 }
208         }
209
210         icp->icmp_code = code;
211         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
212         nip = &icp->icmp_ip;
213
214         /*
215          * Convert fields to network representation.
216          */
217         nip->ip_len = htons(nip->ip_len);
218         nip->ip_off = htons(nip->ip_off);
219
220         /*
221          * Now, copy old ip header (without options)
222          * in front of icmp message.
223          */
224         if (m->m_data - sizeof(struct ip) < m->m_pktdat)
225                 panic("icmp len");
226         m->m_data -= sizeof(struct ip);
227         m->m_len += sizeof(struct ip);
228         m->m_pkthdr.len = m->m_len;
229         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
230         nip = mtod(m, struct ip *);
231         bcopy(oip, nip, sizeof(struct ip));
232         nip->ip_len = m->m_len;
233         nip->ip_vhl = IP_VHL_BORING;
234         nip->ip_p = IPPROTO_ICMP;
235         nip->ip_tos = 0;
236         m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
237         icmp_reflect(m);
238
239 freeit:
240         m_freem(n);
241 }
242
243 /*
244  * Process a received ICMP message.
245  */
246 void
247 icmp_input(struct mbuf *m, ...)
248 {
249         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
250         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
251         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
252         struct icmp *icp;
253         struct in_ifaddr *ia;
254         struct ip *ip = mtod(m, struct ip *);
255         int icmplen = ip->ip_len;
256         int i, hlen;
257         int code, off, proto;
258         __va_list ap;
259
260         __va_start(ap, m);
261         off = __va_arg(ap, int);
262         proto = __va_arg(ap, int);
263         __va_end(ap);
264
265         hlen = off;
266
267         /*
268          * Locate icmp structure in mbuf, and check
269          * that not corrupted and of at least minimum length.
270          */
271 #ifdef ICMPPRINTFS
272         if (icmpprintfs) {
273                 char buf[sizeof "aaa.bbb.ccc.ddd"];
274
275                 strcpy(buf, inet_ntoa(ip->ip_src));
276                 kprintf("icmp_input from %s to %s, len %d\n",
277                        buf, inet_ntoa(ip->ip_dst), icmplen);
278         }
279 #endif
280         if (icmplen < ICMP_MINLEN) {
281                 icmpstat.icps_tooshort++;
282                 goto freeit;
283         }
284         i = hlen + min(icmplen, ICMP_ADVLENMIN);
285         if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
286                 icmpstat.icps_tooshort++;
287                 return;
288         }
289         ip = mtod(m, struct ip *);
290         m->m_len -= hlen;
291         m->m_data += hlen;
292         icp = mtod(m, struct icmp *);
293         if (in_cksum(m, icmplen)) {
294                 icmpstat.icps_checksum++;
295                 goto freeit;
296         }
297         m->m_len += hlen;
298         m->m_data -= hlen;
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                 code = PRC_QUENCH;
383 deliver:
384                 /*
385                  * Problem with datagram; advise higher level routines.
386                  */
387                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
388                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
389                         icmpstat.icps_badlen++;
390                         goto freeit;
391                 }
392                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
393                 /* Discard ICMP's in response to multicast packets */
394                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
395                         goto badcode;
396 #ifdef ICMPPRINTFS
397                 if (icmpprintfs)
398                         kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
399 #endif
400                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
401 #if 1
402                 /*
403                  * MTU discovery:
404                  * If we got a needfrag and there is a host route to the
405                  * original destination, and the MTU is not locked, then
406                  * set the MTU in the route to the suggested new value
407                  * (if given) and then notify as usual.  The ULPs will
408                  * notice that the MTU has changed and adapt accordingly.
409                  * If no new MTU was suggested, then we guess a new one
410                  * less than the current value.  If the new MTU is
411                  * unreasonably small (arbitrarily set at 296), then
412                  * we reset the MTU to the interface value and enable the
413                  * lock bit, indicating that we are no longer doing MTU
414                  * discovery.
415                  */
416                 if (code == PRC_MSGSIZE) {
417                         struct rtentry *rt;
418                         int mtu;
419
420                         rt = rtpurelookup((struct sockaddr *)&icmpsrc);
421                         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
422                             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
423                                 mtu = ntohs(icp->icmp_nextmtu);
424                                 if (!mtu)
425                                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
426                                                           1);
427 #ifdef DEBUG_MTUDISC
428                                 kprintf("MTU for %s reduced to %d\n",
429                                         inet_ntoa(icmpsrc.sin_addr), mtu);
430 #endif
431                                 if (mtu < 296) {
432                                         /* rt->rt_rmx.rmx_mtu =
433                                                 rt->rt_ifp->if_mtu; */
434                                         rt->rt_rmx.rmx_locks |= RTV_MTU;
435                                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
436                                         rt->rt_rmx.rmx_mtu = mtu;
437                                 }
438                         }
439                         if (rt != NULL)
440                                 --rt->rt_refcnt;
441                 }
442 #endif
443                 /*
444                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
445                  * notification to TCP layer.
446                  */
447                 so_pru_ctlinput(
448                         &inetsw[ip_protox[icp->icmp_ip.ip_p]],
449                         code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
450                 break;
451
452 badcode:
453                 icmpstat.icps_badcode++;
454                 break;
455
456         case ICMP_ECHO:
457                 if (!icmpbmcastecho
458                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
459                         icmpstat.icps_bmcastecho++;
460                         break;
461                 }
462                 icp->icmp_type = ICMP_ECHOREPLY;
463 #ifdef ICMP_BANDLIM
464                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
465                         goto freeit;
466                 else
467 #endif
468                         goto reflect;
469
470         case ICMP_TSTAMP:
471                 if (!icmpbmcastecho
472                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
473                         icmpstat.icps_bmcasttstamp++;
474                         break;
475                 }
476                 if (icmplen < ICMP_TSLEN) {
477                         icmpstat.icps_badlen++;
478                         break;
479                 }
480                 icp->icmp_type = ICMP_TSTAMPREPLY;
481                 icp->icmp_rtime = iptime();
482                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
483 #ifdef ICMP_BANDLIM
484                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
485                         goto freeit;
486                 else
487 #endif
488                         goto reflect;
489
490         case ICMP_MASKREQ:
491                 if (icmpmaskrepl == 0)
492                         break;
493                 /*
494                  * We are not able to respond with all ones broadcast
495                  * unless we receive it over a point-to-point interface.
496                  */
497                 if (icmplen < ICMP_MASKLEN)
498                         break;
499                 switch (ip->ip_dst.s_addr) {
500
501                 case INADDR_BROADCAST:
502                 case INADDR_ANY:
503                         icmpdst.sin_addr = ip->ip_src;
504                         break;
505
506                 default:
507                         icmpdst.sin_addr = ip->ip_dst;
508                 }
509                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
510                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
511                 if (ia == 0)
512                         break;
513                 if (ia->ia_ifp == 0)
514                         break;
515                 icp->icmp_type = ICMP_MASKREPLY;
516                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
517                 if (ip->ip_src.s_addr == 0) {
518                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
519                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
520                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
521                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
522                 }
523 reflect:
524                 ip->ip_len += hlen;     /* since ip_input deducts this */
525                 icmpstat.icps_reflect++;
526                 icmpstat.icps_outhist[icp->icmp_type]++;
527                 icmp_reflect(m);
528                 return;
529
530         case ICMP_REDIRECT:
531                 if (log_redirect) {
532                         u_long src, dst, gw;
533
534                         src = ntohl(ip->ip_src.s_addr);
535                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
536                         gw = ntohl(icp->icmp_gwaddr.s_addr);
537                         kprintf("icmp redirect from %d.%d.%d.%d: "
538                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
539                                (int)(src >> 24), (int)((src >> 16) & 0xff),
540                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
541                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
542                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
543                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
544                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
545                 }
546                 if (drop_redirect)
547                         break;
548                 if (code > 3)
549                         goto badcode;
550                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
551                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
552                         icmpstat.icps_badlen++;
553                         break;
554                 }
555                 /*
556                  * Short circuit routing redirects to force
557                  * immediate change in the kernel's routing
558                  * tables.  The message is also handed to anyone
559                  * listening on a raw socket (e.g. the routing
560                  * daemon for use in updating its tables).
561                  */
562                 icmpgw.sin_addr = ip->ip_src;
563                 icmpdst.sin_addr = icp->icmp_gwaddr;
564 #ifdef  ICMPPRINTFS
565                 if (icmpprintfs) {
566                         char buf[sizeof "aaa.bbb.ccc.ddd"];
567
568                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
569                         kprintf("redirect dst %s to %s\n",
570                                buf, inet_ntoa(icp->icmp_gwaddr));
571                 }
572 #endif
573                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
574                 rtredirect((struct sockaddr *)&icmpsrc,
575                   (struct sockaddr *)&icmpdst,
576                   (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
577                   (struct sockaddr *)&icmpgw);
578                 kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
579 #ifdef IPSEC
580                 key_sa_routechange((struct sockaddr *)&icmpsrc);
581 #endif
582                 break;
583
584         /*
585          * No kernel processing for the following;
586          * just fall through to send to raw listener.
587          */
588         case ICMP_ECHOREPLY:
589         case ICMP_ROUTERADVERT:
590         case ICMP_ROUTERSOLICIT:
591         case ICMP_TSTAMPREPLY:
592         case ICMP_IREQREPLY:
593         case ICMP_MASKREPLY:
594         default:
595                 break;
596         }
597
598 raw:
599         rip_input(m, off, proto);
600         return;
601
602 freeit:
603         m_freem(m);
604 }
605
606 /*
607  * Reflect the ip packet back to the source
608  */
609 static void
610 icmp_reflect(struct mbuf *m)
611 {
612         struct ip *ip = mtod(m, struct ip *);
613         struct in_ifaddr *ia;
614         struct in_ifaddr_container *iac;
615         struct in_addr t;
616         struct mbuf *opts = 0;
617         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
618         struct route *ro = NULL, rt;
619
620         if (!in_canforward(ip->ip_src) &&
621             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
622              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
623                 m_freem(m);     /* Bad return address */
624                 icmpstat.icps_badaddr++;
625                 goto done;      /* Ip_output() will check for broadcast */
626         }
627         t = ip->ip_dst;
628         ip->ip_dst = ip->ip_src;
629         ro = &rt;
630         bzero(ro, sizeof *ro);
631         /*
632          * If the incoming packet was addressed directly to us,
633          * use dst as the src for the reply.  Otherwise (broadcast
634          * or anonymous), use the address which corresponds
635          * to the incoming interface.
636          */
637         ia = NULL;
638         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
639                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
640                         ia = iac->ia;
641                         goto match;
642                 }
643         }
644         if (m->m_pkthdr.rcvif != NULL &&
645             m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
646                 struct ifaddr_container *ifac;
647
648                 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
649                               ifa_link) {
650                         struct ifaddr *ifa = ifac->ifa;
651
652                         if (ifa->ifa_addr->sa_family != AF_INET)
653                                 continue;
654                         ia = ifatoia(ifa);
655                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
656                             t.s_addr)
657                                 goto match;
658                 }
659         }
660         ia = ip_rtaddr(ip->ip_dst, ro);
661         /* We need a route to do anything useful. */
662         if (ia == NULL) {
663                 m_freem(m);
664                 icmpstat.icps_noroute++;
665                 goto done;
666         }
667 match:
668         t = IA_SIN(ia)->sin_addr;
669         ip->ip_src = t;
670         ip->ip_ttl = ip_defttl;
671
672         if (optlen > 0) {
673                 u_char *cp;
674                 int opt, cnt;
675                 u_int len;
676
677                 /*
678                  * Retrieve any source routing from the incoming packet;
679                  * add on any record-route or timestamp options.
680                  */
681                 cp = (u_char *) (ip + 1);
682                 if ((opts = ip_srcroute(m)) == 0 &&
683                     (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
684                         opts->m_len = sizeof(struct in_addr);
685                         mtod(opts, struct in_addr *)->s_addr = 0;
686                 }
687                 if (opts) {
688 #ifdef ICMPPRINTFS
689                         if (icmpprintfs)
690                                 kprintf("icmp_reflect optlen %d rt %d => ",
691                                        optlen, opts->m_len);
692 #endif
693                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
694                                 opt = cp[IPOPT_OPTVAL];
695                                 if (opt == IPOPT_EOL)
696                                         break;
697                                 if (opt == IPOPT_NOP)
698                                         len = 1;
699                                 else {
700                                         if (cnt < IPOPT_OLEN + sizeof *cp)
701                                                 break;
702                                         len = cp[IPOPT_OLEN];
703                                         if (len < IPOPT_OLEN + sizeof *cp ||
704                                             len > cnt)
705                                         break;
706                                 }
707                                 /*
708                                  * Should check for overflow, but it
709                                  * "can't happen".
710                                  */
711                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
712                                     opt == IPOPT_SECURITY) {
713                                         bcopy(cp,
714                                               mtod(opts, caddr_t) + opts->m_len,
715                                               len);
716                                         opts->m_len += len;
717                                 }
718                         }
719                         /* Terminate & pad, if necessary */
720                         cnt = opts->m_len % 4;
721                         if (cnt) {
722                                 for (; cnt < 4; cnt++) {
723                                         *(mtod(opts, caddr_t) + opts->m_len) =
724                                             IPOPT_EOL;
725                                         opts->m_len++;
726                                 }
727                         }
728 #ifdef ICMPPRINTFS
729                         if (icmpprintfs)
730                                 kprintf("%d\n", opts->m_len);
731 #endif
732                 }
733                 /*
734                  * Now strip out original options by copying rest of first
735                  * mbuf's data back, and adjust the IP length.
736                  */
737                 ip->ip_len -= optlen;
738                 ip->ip_vhl = IP_VHL_BORING;
739                 m->m_len -= optlen;
740                 if (m->m_flags & M_PKTHDR)
741                         m->m_pkthdr.len -= optlen;
742                 optlen += sizeof(struct ip);
743                 bcopy((caddr_t)ip + optlen, ip + 1,
744                       m->m_len - sizeof(struct ip));
745         }
746         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
747         m->m_flags &= ~(M_BCAST|M_MCAST);
748         icmp_send(m, opts, ro);
749 done:
750         if (opts)
751                 m_free(opts);
752         if (ro && ro->ro_rt)
753                 RTFREE(ro->ro_rt);
754 }
755
756 /*
757  * Send an icmp packet back to the ip level,
758  * after supplying a checksum.
759  */
760 static void
761 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
762 {
763         struct ip *ip = mtod(m, struct ip *);
764         int hlen;
765         struct icmp *icp;
766
767         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
768         m->m_data += hlen;
769         m->m_len -= hlen;
770         icp = mtod(m, struct icmp *);
771         icp->icmp_cksum = 0;
772         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
773         m->m_data -= hlen;
774         m->m_len += hlen;
775         m->m_pkthdr.rcvif = (struct ifnet *)0;
776 #ifdef ICMPPRINTFS
777         if (icmpprintfs) {
778                 char buf[sizeof "aaa.bbb.ccc.ddd"];
779
780                 strcpy(buf, inet_ntoa(ip->ip_dst));
781                 kprintf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src));
782         }
783 #endif
784         ip_output(m, opts, rt, 0, NULL, NULL);
785 }
786
787 n_time
788 iptime(void)
789 {
790         struct timeval atv;
791         u_long t;
792
793         getmicrotime(&atv);
794         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
795         return (htonl(t));
796 }
797
798 #if 1
799 /*
800  * Return the next larger or smaller MTU plateau (table from RFC 1191)
801  * given current value MTU.  If DIR is less than zero, a larger plateau
802  * is returned; otherwise, a smaller value is returned.
803  */
804 int
805 ip_next_mtu(int mtu, int dir)
806 {
807         static int mtutab[] = {
808                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
809                 68, 0
810         };
811         int i;
812
813         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
814                 if (mtu >= mtutab[i])
815                         break;
816         }
817
818         if (dir < 0) {
819                 if (i == 0) {
820                         return 0;
821                 } else {
822                         return mtutab[i - 1];
823                 }
824         } else {
825                 if (mtutab[i] == 0) {
826                         return 0;
827                 } else if(mtu > mtutab[i]) {
828                         return mtutab[i];
829                 } else {
830                         return mtutab[i + 1];
831                 }
832         }
833 }
834 #endif
835
836 #ifdef ICMP_BANDLIM
837 /*
838  * badport_bandlim() - check for ICMP bandwidth limit
839  *
840  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
841  *      hit our bandwidth limit and it is not ok.
842  *
843  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
844  *
845  *      For now we separate the TCP and UDP subsystems w/ different 'which'
846  *      values.  We may eventually remove this separation (and simplify the
847  *      code further).
848  *
849  *      Note that the printing of the error message is delayed so we can
850  *      properly print the icmp error rate that the system was trying to do
851  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
852  *      the 'final' error, but it doesn't make sense to solve the printing
853  *      delay with more complex code.
854  */
855 int
856 badport_bandlim(int which)
857 {
858         static int lticks[BANDLIM_MAX + 1];
859         static int lpackets[BANDLIM_MAX + 1];
860         int dticks;
861         const char *bandlimittype[] = {
862                 "Limiting icmp unreach response",
863                 "Limiting icmp ping response",
864                 "Limiting icmp tstamp response",
865                 "Limiting closed port RST response",
866                 "Limiting open port RST response"
867                 };
868
869         /*
870          * Return ok status if feature disabled or argument out of
871          * ranage.
872          */
873
874         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
875                 return(0);
876         dticks = ticks - lticks[which];
877
878         /*
879          * reset stats when cumulative dt exceeds one second.
880          */
881
882         if ((unsigned int)dticks > hz) {
883                 if (lpackets[which] > icmplim && icmplim_output) {
884                         kprintf("%s from %d to %d packets per second\n",
885                                 bandlimittype[which],
886                                 lpackets[which],
887                                 icmplim
888                         );
889                 }
890                 lticks[which] = ticks;
891                 lpackets[which] = 0;
892         }
893
894         /*
895          * bump packet count
896          */
897
898         if (++lpackets[which] > icmplim) {
899                 return(-1);
900         }
901         return(0);
902 }
903 #endif