projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[IPV6]: Make struct ip6_prohibit_entry_template static.
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv4 / icmp.c
CommitLineData
1da177e4
LT		 1/*
2 * NET3: Implementation of the ICMP protocol layer.
3 *
4 * Alan Cox, <alan@redhat.com>
5 *
6 * Version: $Id: icmp.c,v 1.85 2002/02/01 22:01:03 davem Exp $
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 *
13 * Some of the function names and the icmp unreach table for this
14 * module were derived from [icmp.c 1.0.11 06/02/93] by
15 * Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
16 * Other than that this module is a complete rewrite.
17 *
18 * Fixes:
19 * Clemens Fruhwirth : introduce global icmp rate limiting
20 * with icmp type masking ability instead
21 * of broken per type icmp timeouts.
22 * Mike Shaver : RFC1122 checks.
23 * Alan Cox : Multicast ping reply as self.
24 * Alan Cox : Fix atomicity lockup in ip_build_xmit
25 * call.
26 * Alan Cox : Added 216,128 byte paths to the MTU
27 * code.
28 * Martin Mares : RFC1812 checks.
29 * Martin Mares : Can be configured to follow redirects
30 * if acting as a router _without_ a
31 * routing protocol (RFC 1812).
32 * Martin Mares : Echo requests may be configured to
33 * be ignored (RFC 1812).
34 * Martin Mares : Limitation of ICMP error message
35 * transmit rate (RFC 1812).
36 * Martin Mares : TOS and Precedence set correctly
37 * (RFC 1812).
38 * Martin Mares : Now copying as much data from the
39 * original packet as we can without
40 * exceeding 576 bytes (RFC 1812).
41 * Willy Konynenberg : Transparent proxying support.
42 * Keith Owens : RFC1191 correction for 4.2BSD based
43 * path MTU bug.
44 * Thomas Quinot : ICMP Dest Unreach codes up to 15 are
45 * valid (RFC 1812).
46 * Andi Kleen : Check all packet lengths properly
47 * and moved all kfree_skb() up to
48 * icmp_rcv.
49 * Andi Kleen : Move the rate limit bookkeeping
50 * into the dest entry and use a token
51 * bucket filter (thanks to ANK). Make
52 * the rates sysctl configurable.
53 * Yu Tianli : Fixed two ugly bugs in icmp_send
54 * - IP option length was accounted wrongly
55 * - ICMP header length was not accounted
56 * at all.
57 * Tristan Greaves : Added sysctl option to ignore bogus
58 * broadcast responses from broken routers.
59 *
60 * To Fix:
61 *
62 * - Should use skb_pull() instead of all the manual checking.
63 * This would also greatly simply some upper layer error handlers. --AK
64 *
65 */
66
1da177e4
LT		 67#include <linux/module.h>
68#include <linux/types.h>
69#include <linux/jiffies.h>
70#include <linux/kernel.h>
71#include <linux/fcntl.h>
72#include <linux/socket.h>
73#include <linux/in.h>
74#include <linux/inet.h>
14c85021 75#include <linux/inetdevice.h>
1da177e4
LT		 76#include <linux/netdevice.h>
77#include <linux/string.h>
78#include <linux/netfilter_ipv4.h>
79#include <net/snmp.h>
80#include <net/ip.h>
81#include <net/route.h>
82#include <net/protocol.h>
83#include <net/icmp.h>
84#include <net/tcp.h>
85#include <net/udp.h>
86#include <net/raw.h>
87#include <linux/skbuff.h>
88#include <net/sock.h>
89#include <linux/errno.h>
90#include <linux/timer.h>
91#include <linux/init.h>
92#include <asm/system.h>
93#include <asm/uaccess.h>
94#include <net/checksum.h>
8b7817f3 95#include <net/xfrm.h>
c1e9894d 96#include <net/inet_common.h>
1da177e4
LT		 97
98/*
99 * Build xmit assembly blocks
100 */
101
102struct icmp_bxm {
103 struct sk_buff *skb;
104 int offset;
105 int data_len;
106
107 struct {
108 struct icmphdr icmph;
b03d73e3 109 __be32 times[3];
1da177e4
LT		 110 } data;
111 int head_len;
112 struct ip_options replyopts;
113 unsigned char optbuf[40];
114};
115
116/*
117 * Statistics
118 */
ba89966c 119DEFINE_SNMP_STAT(struct icmp_mib, icmp_statistics) __read_mostly;
96793b48 120DEFINE_SNMP_STAT(struct icmpmsg_mib, icmpmsg_statistics) __read_mostly;
1da177e4
LT		 121
122/* An array of errno for error messages from dest unreach. */
123/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
124
125struct icmp_err icmp_err_convert[] = {
126 {
127 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
128 .fatal = 0,
129 },
130 {
131 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
132 .fatal = 0,
133 },
134 {
135 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
136 .fatal = 1,
137 },
138 {
139 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
140 .fatal = 1,
141 },
142 {
143 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
144 .fatal = 0,
145 },
146 {
147 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
148 .fatal = 0,
149 },
150 {
151 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
152 .fatal = 1,
153 },
154 {
155 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
156 .fatal = 1,
157 },
158 {
159 .errno = ENONET, /* ICMP_HOST_ISOLATED */
160 .fatal = 1,
161 },
162 {
163 .errno = ENETUNREACH, /* ICMP_NET_ANO */
164 .fatal = 1,
165 },
166 {
167 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
168 .fatal = 1,
169 },
170 {
171 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
172 .fatal = 0,
173 },
174 {
175 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
176 .fatal = 0,
177 },
178 {
179 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
180 .fatal = 1,
181 },
182 {
183 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
184 .fatal = 1,
185 },
186 {
187 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
188 .fatal = 1,
189 },
190};
191
1da177e4
LT		 192/*
193 * ICMP control array. This specifies what to do with each ICMP.
194 */
195
196struct icmp_control {
1da177e4
LT		 197 void (*handler)(struct sk_buff *skb);
198 short error; /* This ICMP is classed as an error message */
199};
200
9b5b5cff 201static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
1da177e4
LT		 202
203/*
204 * The ICMP socket(s). This is the most convenient way to flow control
205 * our ICMP output as well as maintain a clean interface throughout
206 * all layers. All Socketless IP sends will soon be gone.
207 *
208 * On SMP we have one ICMP socket per-cpu.
209 */
4a6ad7a1
DL		 210static struct sock *icmp_sk(struct net *net)
211{
212 return net->ipv4.icmp_sk[smp_processor_id()];
213}
1da177e4 214
405666db 215static inline int icmp_xmit_lock(struct sock *sk)
1da177e4
LT		 216{
217 local_bh_disable();
218
405666db 219 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
1da177e4
LT		 220 /* This can happen if the output path signals a
221 * dst_link_failure() for an outgoing ICMP packet.
222 */
223 local_bh_enable();
224 return 1;
225 }
226 return 0;
227}
228
405666db 229static inline void icmp_xmit_unlock(struct sock *sk)
1da177e4 230{
405666db 231 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 232}
233
234/*
235 * Send an ICMP frame.
236 */
237
238/*
239 * Check transmit rate limitation for given message.
240 * The rate information is held in the destination cache now.
241 * This function is generic and could be used for other purposes
242 * too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
243 *
244 * Note that the same dst_entry fields are modified by functions in
245 * route.c too, but these work for packet destinations while xrlim_allow
246 * works for icmp destinations. This means the rate limiting information
247 * for one "ip object" is shared - and these ICMPs are twice limited:
248 * by source and by destination.
249 *
250 * RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
251 * SHOULD allow setting of rate limits
252 *
253 * Shared between ICMPv4 and ICMPv6.
254 */
255#define XRLIM_BURST_FACTOR 6
256int xrlim_allow(struct dst_entry *dst, int timeout)
257{
69a73829 258 unsigned long now, token = dst->rate_tokens;
1da177e4
LT		 259 int rc = 0;
260
261 now = jiffies;
69a73829 262 token += now - dst->rate_last;
1da177e4 263 dst->rate_last = now;
69a73829
ED		 264 if (token > XRLIM_BURST_FACTOR * timeout)
265 token = XRLIM_BURST_FACTOR * timeout;
266 if (token >= timeout) {
267 token -= timeout;
1da177e4
LT		 268 rc = 1;
269 }
69a73829 270 dst->rate_tokens = token;
1da177e4
LT		 271 return rc;
272}
273
b34a95ee
PE		 274static inline int icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
275 int type, int code)
1da177e4
LT		 276{
277 struct dst_entry *dst = &rt->u.dst;
278 int rc = 1;
279
280 if (type > NR_ICMP_TYPES)
281 goto out;
282
283 /* Don't limit PMTU discovery. */
284 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
285 goto out;
286
287 /* No rate limit on loopback */
288 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
e905a9ed 289 goto out;
1da177e4
LT		 290
291 /* Limit if icmp type is enabled in ratemask. */
b34a95ee
PE		 292 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask)
293 rc = xrlim_allow(dst, net->ipv4.sysctl_icmp_ratelimit);
1da177e4
LT		 294out:
295 return rc;
296}
297
298/*
299 * Maintain the counters used in the SNMP statistics for outgoing ICMP
300 */
96793b48 301void icmp_out_count(unsigned char type)
1da177e4 302{
96793b48
DS		 303 ICMPMSGOUT_INC_STATS(type);
304 ICMP_INC_STATS(ICMP_MIB_OUTMSGS);
1da177e4
LT		 305}
306
307/*
308 * Checksum each fragment, and on the first include the headers and final
309 * checksum.
310 */
311static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
312 struct sk_buff *skb)
313{
314 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
5f92a738 315 __wsum csum;
1da177e4
LT		 316
317 csum = skb_copy_and_csum_bits(icmp_param->skb,
318 icmp_param->offset + offset,
319 to, len, 0);
320
321 skb->csum = csum_block_add(skb->csum, csum, odd);
322 if (icmp_pointers[icmp_param->data.icmph.type].error)
323 nf_ct_attach(skb, icmp_param->skb);
324 return 0;
325}
326
327static void icmp_push_reply(struct icmp_bxm *icmp_param,
328 struct ipcm_cookie *ipc, struct rtable *rt)
329{
1e3cf683 330 struct sock *sk;
1da177e4
LT		 331 struct sk_buff *skb;
332
c346dca1 333 sk = icmp_sk(dev_net(rt->u.dst.dev));
1e3cf683 334 if (ip_append_data(sk, icmp_glue_bits, icmp_param,
e905a9ed
YH		 335 icmp_param->data_len+icmp_param->head_len,
336 icmp_param->head_len,
337 ipc, rt, MSG_DONTWAIT) < 0)
1e3cf683
DL		 338 ip_flush_pending_frames(sk);
339 else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
88c7664f 340 struct icmphdr *icmph = icmp_hdr(skb);
d3bc23e7 341 __wsum csum = 0;
1da177e4
LT		 342 struct sk_buff *skb1;
343
1e3cf683 344 skb_queue_walk(&sk->sk_write_queue, skb1) {
1da177e4
LT		 345 csum = csum_add(csum, skb1->csum);
346 }
347 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
348 (char *)icmph,
349 icmp_param->head_len, csum);
350 icmph->checksum = csum_fold(csum);
351 skb->ip_summed = CHECKSUM_NONE;
1e3cf683 352 ip_push_pending_frames(sk);
1da177e4
LT		 353 }
354}
355
356/*
357 * Driving logic for building and sending ICMP messages.
358 */
359
360static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
361{
1da177e4 362 struct ipcm_cookie ipc;
ee6b9673 363 struct rtable *rt = skb->rtable;
c346dca1 364 struct net *net = dev_net(rt->u.dst.dev);
4a6ad7a1
DL		 365 struct sock *sk = icmp_sk(net);
366 struct inet_sock *inet = inet_sk(sk);
3ca3c68e 367 __be32 daddr;
1da177e4
LT		 368
369 if (ip_options_echo(&icmp_param->replyopts, skb))
f00c401b 370 return;
1da177e4 371
405666db 372 if (icmp_xmit_lock(sk))
1da177e4
LT		 373 return;
374
375 icmp_param->data.icmph.checksum = 0;
1da177e4 376
eddc9ec5 377 inet->tos = ip_hdr(skb)->tos;
1da177e4
LT		 378 daddr = ipc.addr = rt->rt_src;
379 ipc.opt = NULL;
380 if (icmp_param->replyopts.optlen) {
381 ipc.opt = &icmp_param->replyopts;
382 if (ipc.opt->srr)
383 daddr = icmp_param->replyopts.faddr;
384 }
385 {
386 struct flowi fl = { .nl_u = { .ip4_u =
387 { .daddr = daddr,
388 .saddr = rt->rt_spec_dst,
eddc9ec5 389 .tos = RT_TOS(ip_hdr(skb)->tos) } },
1da177e4 390 .proto = IPPROTO_ICMP };
beb8d13b 391 security_skb_classify_flow(skb, &fl);
4a6ad7a1 392 if (ip_route_output_key(net, &rt, &fl))
1da177e4
LT		 393 goto out_unlock;
394 }
b34a95ee 395 if (icmpv4_xrlim_allow(net, rt, icmp_param->data.icmph.type,
1da177e4
LT		 396 icmp_param->data.icmph.code))
397 icmp_push_reply(icmp_param, &ipc, rt);
398 ip_rt_put(rt);
399out_unlock:
405666db 400 icmp_xmit_unlock(sk);
1da177e4
LT		 401}
402
403
404/*
405 * Send an ICMP message in response to a situation
406 *
407 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
408 * MAY send more (we do).
409 * MUST NOT change this header information.
410 * MUST NOT reply to a multicast/broadcast IP address.
411 * MUST NOT reply to a multicast/broadcast MAC address.
412 * MUST reply to only the first fragment.
413 */
414
e4883014 415void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
1da177e4
LT		 416{
417 struct iphdr *iph;
418 int room;
419 struct icmp_bxm icmp_param;
ee6b9673 420 struct rtable *rt = skb_in->rtable;
1da177e4 421 struct ipcm_cookie ipc;
a61ced5d 422 __be32 saddr;
1da177e4 423 u8 tos;
dde1bc0e 424 struct net *net;
4a6ad7a1 425 struct sock *sk;
1da177e4
LT		 426
427 if (!rt)
428 goto out;
c346dca1 429 net = dev_net(rt->u.dst.dev);
4a6ad7a1 430 sk = icmp_sk(net);
1da177e4
LT		 431
432 /*
433 * Find the original header. It is expected to be valid, of course.
434 * Check this, icmp_send is called from the most obscure devices
435 * sometimes.
436 */
eddc9ec5 437 iph = ip_hdr(skb_in);
1da177e4 438
27a884dc
ACM		 439 if ((u8 *)iph < skb_in->head ||
440 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
1da177e4
LT		 441 goto out;
442
443 /*
444 * No replies to physical multicast/broadcast
445 */
446 if (skb_in->pkt_type != PACKET_HOST)
447 goto out;
448
449 /*
450 * Now check at the protocol level
451 */
452 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
453 goto out;
454
455 /*
456 * Only reply to fragment 0. We byte re-order the constant
457 * mask for efficiency.
458 */
459 if (iph->frag_off & htons(IP_OFFSET))
460 goto out;
461
462 /*
463 * If we send an ICMP error to an ICMP error a mess would result..
464 */
465 if (icmp_pointers[type].error) {
466 /*
467 * We are an error, check if we are replying to an
468 * ICMP error
469 */
470 if (iph->protocol == IPPROTO_ICMP) {
471 u8 _inner_type, *itp;
472
473 itp = skb_header_pointer(skb_in,
d56f90a7 474 skb_network_header(skb_in) +
1da177e4
LT		 475 (iph->ihl << 2) +
476 offsetof(struct icmphdr,
477 type) -
478 skb_in->data,
479 sizeof(_inner_type),
480 &_inner_type);
481 if (itp == NULL)
482 goto out;
483
484 /*
485 * Assume any unknown ICMP type is an error. This
486 * isn't specified by the RFC, but think about it..
487 */
488 if (*itp > NR_ICMP_TYPES ||
489 icmp_pointers[*itp].error)
490 goto out;
491 }
492 }
493
405666db 494 if (icmp_xmit_lock(sk))
1da177e4
LT		 495 return;
496
497 /*
498 * Construct source address and options.
499 */
500
501 saddr = iph->daddr;
1c2fb7f9 502 if (!(rt->rt_flags & RTCF_LOCAL)) {
6e1d9103
PM		 503 struct net_device *dev = NULL;
504
a24022e1 505 if (rt->fl.iif &&
b34a95ee 506 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
dde1bc0e 507 dev = dev_get_by_index(net, rt->fl.iif);
6e1d9103
PM		 508
509 if (dev) {
510 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
511 dev_put(dev);
512 } else
1c2fb7f9
S		 513 saddr = 0;
514 }
1da177e4
LT		 515
516 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
517 IPTOS_PREC_INTERNETCONTROL) :
518 iph->tos;
519
520 if (ip_options_echo(&icmp_param.replyopts, skb_in))
fa60cf7f 521 goto out_unlock;
1da177e4
LT		 522
523
524 /*
525 * Prepare data for ICMP header.
526 */
527
528 icmp_param.data.icmph.type = type;
529 icmp_param.data.icmph.code = code;
530 icmp_param.data.icmph.un.gateway = info;
531 icmp_param.data.icmph.checksum = 0;
532 icmp_param.skb = skb_in;
d56f90a7 533 icmp_param.offset = skb_network_offset(skb_in);
405666db 534 inet_sk(sk)->tos = tos;
1da177e4
LT		 535 ipc.addr = iph->saddr;
536 ipc.opt = &icmp_param.replyopts;
537
538 {
539 struct flowi fl = {
540 .nl_u = {
541 .ip4_u = {
542 .daddr = icmp_param.replyopts.srr ?
543 icmp_param.replyopts.faddr :
544 iph->saddr,
545 .saddr = saddr,
546 .tos = RT_TOS(tos)
547 }
548 },
549 .proto = IPPROTO_ICMP,
550 .uli_u = {
551 .icmpt = {
552 .type = type,
553 .code = code
554 }
555 }
556 };
8b7817f3
HX		 557 int err;
558 struct rtable *rt2;
559
beb8d13b 560 security_skb_classify_flow(skb_in, &fl);
dde1bc0e 561 if (__ip_route_output_key(net, &rt, &fl))
8b7817f3
HX		 562 goto out_unlock;
563
564 /* No need to clone since we're just using its address. */
565 rt2 = rt;
566
567 err = xfrm_lookup((struct dst_entry **)&rt, &fl, NULL, 0);
568 switch (err) {
569 case 0:
570 if (rt != rt2)
571 goto route_done;
572 break;
573 case -EPERM:
574 rt = NULL;
575 break;
576 default:
577 goto out_unlock;
578 }
579
580 if (xfrm_decode_session_reverse(skb_in, &fl, AF_INET))
af268182 581 goto relookup_failed;
8b7817f3 582
dde1bc0e
DL		 583 if (inet_addr_type(net, fl.fl4_src) == RTN_LOCAL)
584 err = __ip_route_output_key(net, &rt2, &fl);
8b7817f3
HX		 585 else {
586 struct flowi fl2 = {};
587 struct dst_entry *odst;
588
589 fl2.fl4_dst = fl.fl4_src;
dde1bc0e 590 if (ip_route_output_key(net, &rt2, &fl2))
af268182 591 goto relookup_failed;
8b7817f3
HX		 592
593 /* Ugh! */
594 odst = skb_in->dst;
595 err = ip_route_input(skb_in, fl.fl4_dst, fl.fl4_src,
596 RT_TOS(tos), rt2->u.dst.dev);
597
598 dst_release(&rt2->u.dst);
ee6b9673 599 rt2 = skb_in->rtable;
8b7817f3
HX		 600 skb_in->dst = odst;
601 }
602
603 if (err)
af268182 604 goto relookup_failed;
8b7817f3
HX		 605
606 err = xfrm_lookup((struct dst_entry **)&rt2, &fl, NULL,
607 XFRM_LOOKUP_ICMP);
af268182
HX		 608 switch (err) {
609 case 0:
610 dst_release(&rt->u.dst);
611 rt = rt2;
612 break;
613 case -EPERM:
614 goto ende;
615 default:
616relookup_failed:
8b7817f3
HX		 617 if (!rt)
618 goto out_unlock;
af268182 619 break;
8b7817f3 620 }
1da177e4
LT		 621 }
622
8b7817f3 623route_done:
b34a95ee 624 if (!icmpv4_xrlim_allow(net, rt, type, code))
1da177e4
LT		 625 goto ende;
626
627 /* RFC says return as much as we can without exceeding 576 bytes. */
628
629 room = dst_mtu(&rt->u.dst);
630 if (room > 576)
631 room = 576;
632 room -= sizeof(struct iphdr) + icmp_param.replyopts.optlen;
633 room -= sizeof(struct icmphdr);
634
635 icmp_param.data_len = skb_in->len - icmp_param.offset;
636 if (icmp_param.data_len > room)
637 icmp_param.data_len = room;
638 icmp_param.head_len = sizeof(struct icmphdr);
639
640 icmp_push_reply(&icmp_param, &ipc, rt);
641ende:
642 ip_rt_put(rt);
643out_unlock:
405666db 644 icmp_xmit_unlock(sk);
1da177e4
LT		 645out:;
646}
647
648
649/*
650 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
651 */
652
653static void icmp_unreach(struct sk_buff *skb)
654{
655 struct iphdr *iph;
656 struct icmphdr *icmph;
657 int hash, protocol;
658 struct net_protocol *ipprot;
1da177e4 659 u32 info = 0;
dde1bc0e
DL		 660 struct net *net;
661
c346dca1 662 net = dev_net(skb->dst->dev);
1da177e4
LT		 663
664 /*
665 * Incomplete header ?
666 * Only checks for the IP header, there should be an
667 * additional check for longer headers in upper levels.
668 */
669
670 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
671 goto out_err;
672
88c7664f 673 icmph = icmp_hdr(skb);
1da177e4
LT		 674 iph = (struct iphdr *)skb->data;
675
676 if (iph->ihl < 5) /* Mangled header, drop. */
677 goto out_err;
678
679 if (icmph->type == ICMP_DEST_UNREACH) {
680 switch (icmph->code & 15) {
681 case ICMP_NET_UNREACH:
682 case ICMP_HOST_UNREACH:
683 case ICMP_PROT_UNREACH:
684 case ICMP_PORT_UNREACH:
685 break;
686 case ICMP_FRAG_NEEDED:
687 if (ipv4_config.no_pmtu_disc) {
a7d632b6 688 LIMIT_NETDEBUG(KERN_INFO "ICMP: " NIPQUAD_FMT ": "
1da177e4
LT		 689 "fragmentation needed "
690 "and DF set.\n",
64ce2073 691 NIPQUAD(iph->daddr));
1da177e4 692 } else {
dde1bc0e 693 info = ip_rt_frag_needed(net, iph,
1da177e4
LT		 694 ntohs(icmph->un.frag.mtu));
695 if (!info)
696 goto out;
697 }
698 break;
699 case ICMP_SR_FAILED:
a7d632b6 700 LIMIT_NETDEBUG(KERN_INFO "ICMP: " NIPQUAD_FMT ": Source "
1da177e4 701 "Route Failed.\n",
64ce2073 702 NIPQUAD(iph->daddr));
1da177e4
LT		 703 break;
704 default:
705 break;
706 }
707 if (icmph->code > NR_ICMP_UNREACH)
708 goto out;
709 } else if (icmph->type == ICMP_PARAMETERPROB)
710 info = ntohl(icmph->un.gateway) >> 24;
711
712 /*
713 * Throw it at our lower layers
714 *
715 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
716 * header.
717 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
718 * transport layer.
719 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
720 * transport layer.
721 */
722
723 /*
724 * Check the other end isnt violating RFC 1122. Some routers send
725 * bogus responses to broadcast frames. If you see this message
726 * first check your netmask matches at both ends, if it does then
727 * get the other vendor to fix their kit.
728 */
729
b34a95ee 730 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
dde1bc0e 731 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
1da177e4 732 if (net_ratelimit())
a7d632b6 733 printk(KERN_WARNING NIPQUAD_FMT " sent an invalid ICMP "
1da177e4 734 "type %u, code %u "
a7d632b6 735 "error to a broadcast: " NIPQUAD_FMT " on %s\n",
eddc9ec5 736 NIPQUAD(ip_hdr(skb)->saddr),
1da177e4
LT		 737 icmph->type, icmph->code,
738 NIPQUAD(iph->daddr),
739 skb->dev->name);
740 goto out;
741 }
742
743 /* Checkin full IP header plus 8 bytes of protocol to
744 * avoid additional coding at protocol handlers.
745 */
746 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
747 goto out;
748
749 iph = (struct iphdr *)skb->data;
750 protocol = iph->protocol;
751
752 /*
753 * Deliver ICMP message to raw sockets. Pretty useless feature?
754 */
7bc54c90 755 raw_icmp_error(skb, protocol, info);
1da177e4 756
1da177e4 757 hash = protocol & (MAX_INET_PROTOS - 1);
1da177e4
LT		 758 rcu_read_lock();
759 ipprot = rcu_dereference(inet_protos[hash]);
760 if (ipprot && ipprot->err_handler)
761 ipprot->err_handler(skb, info);
762 rcu_read_unlock();
763
764out:
765 return;
766out_err:
767 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
768 goto out;
769}
770
771
772/*
773 * Handle ICMP_REDIRECT.
774 */
775
776static void icmp_redirect(struct sk_buff *skb)
777{
778 struct iphdr *iph;
1da177e4
LT		 779
780 if (skb->len < sizeof(struct iphdr))
781 goto out_err;
782
783 /*
784 * Get the copied header of the packet that caused the redirect
785 */
786 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
787 goto out;
788
789 iph = (struct iphdr *)skb->data;
1da177e4 790
88c7664f 791 switch (icmp_hdr(skb)->code & 7) {
1da177e4
LT		 792 case ICMP_REDIR_NET:
793 case ICMP_REDIR_NETTOS:
794 /*
795 * As per RFC recommendations now handle it as a host redirect.
796 */
797 case ICMP_REDIR_HOST:
798 case ICMP_REDIR_HOSTTOS:
eddc9ec5 799 ip_rt_redirect(ip_hdr(skb)->saddr, iph->daddr,
88c7664f 800 icmp_hdr(skb)->un.gateway,
cef2685e 801 iph->saddr, skb->dev);
1da177e4 802 break;
e905a9ed 803 }
1da177e4
LT		 804out:
805 return;
806out_err:
807 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
808 goto out;
809}
810
811/*
812 * Handle ICMP_ECHO ("ping") requests.
813 *
814 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
815 * requests.
816 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
817 * included in the reply.
818 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
819 * echo requests, MUST have default=NOT.
820 * See also WRT handling of options once they are done and working.
821 */
822
823static void icmp_echo(struct sk_buff *skb)
824{
b34a95ee
PE		 825 struct net *net;
826
789e41e6 827 net = dev_net(skb->dst->dev);
b34a95ee 828 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
1da177e4
LT		 829 struct icmp_bxm icmp_param;
830
88c7664f 831 icmp_param.data.icmph = *icmp_hdr(skb);
1da177e4
LT		 832 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
833 icmp_param.skb = skb;
834 icmp_param.offset = 0;
835 icmp_param.data_len = skb->len;
836 icmp_param.head_len = sizeof(struct icmphdr);
837 icmp_reply(&icmp_param, skb);
838 }
839}
840
841/*
842 * Handle ICMP Timestamp requests.
843 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
844 * SHOULD be in the kernel for minimum random latency.
845 * MUST be accurate to a few minutes.
846 * MUST be updated at least at 15Hz.
847 */
848static void icmp_timestamp(struct sk_buff *skb)
849{
850 struct timeval tv;
851 struct icmp_bxm icmp_param;
852 /*
853 * Too short.
854 */
855 if (skb->len < 4)
856 goto out_err;
857
858 /*
859 * Fill in the current time as ms since midnight UT:
860 */
861 do_gettimeofday(&tv);
862 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * 1000 +
863 tv.tv_usec / 1000);
864 icmp_param.data.times[2] = icmp_param.data.times[1];
865 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
866 BUG();
88c7664f 867 icmp_param.data.icmph = *icmp_hdr(skb);
1da177e4
LT		 868 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
869 icmp_param.data.icmph.code = 0;
870 icmp_param.skb = skb;
871 icmp_param.offset = 0;
872 icmp_param.data_len = 0;
873 icmp_param.head_len = sizeof(struct icmphdr) + 12;
874 icmp_reply(&icmp_param, skb);
875out:
876 return;
877out_err:
878 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
879 goto out;
880}
881
882
883/*
884 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
885 *
886 * RFC1122 (3.2.2.9). A host MUST only send replies to
887 * ADDRESS_MASK requests if it's been configured as an address mask
888 * agent. Receiving a request doesn't constitute implicit permission to
889 * act as one. Of course, implementing this correctly requires (SHOULD)
890 * a way to turn the functionality on and off. Another one for sysctl(),
891 * I guess. -- MS
892 *
893 * RFC1812 (4.3.3.9). A router MUST implement it.
894 * A router SHOULD have switch turning it on/off.
895 * This switch MUST be ON by default.
896 *
897 * Gratuitous replies, zero-source replies are not implemented,
898 * that complies with RFC. DO NOT implement them!!! All the idea
899 * of broadcast addrmask replies as specified in RFC950 is broken.
900 * The problem is that it is not uncommon to have several prefixes
901 * on one physical interface. Moreover, addrmask agent can even be
902 * not aware of existing another prefixes.
903 * If source is zero, addrmask agent cannot choose correct prefix.
904 * Gratuitous mask announcements suffer from the same problem.
905 * RFC1812 explains it, but still allows to use ADDRMASK,
906 * that is pretty silly. --ANK
907 *
908 * All these rules are so bizarre, that I removed kernel addrmask
909 * support at all. It is wrong, it is obsolete, nobody uses it in
910 * any case. --ANK
911 *
912 * Furthermore you can do it with a usermode address agent program
913 * anyway...
914 */
915
916static void icmp_address(struct sk_buff *skb)
917{
918#if 0
919 if (net_ratelimit())
920 printk(KERN_DEBUG "a guy asks for address mask. Who is it?\n");
921#endif
922}
923
924/*
925 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
926 * loudly if an inconsistency is found.
927 */
928
929static void icmp_address_reply(struct sk_buff *skb)
930{
ee6b9673 931 struct rtable *rt = skb->rtable;
1da177e4
LT		 932 struct net_device *dev = skb->dev;
933 struct in_device *in_dev;
934 struct in_ifaddr *ifa;
935
936 if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC))
937 goto out;
938
939 in_dev = in_dev_get(dev);
940 if (!in_dev)
941 goto out;
942 rcu_read_lock();
943 if (in_dev->ifa_list &&
944 IN_DEV_LOG_MARTIANS(in_dev) &&
945 IN_DEV_FORWARD(in_dev)) {
a144ea4b 946 __be32 _mask, *mp;
1da177e4
LT		 947
948 mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask);
09a62660 949 BUG_ON(mp == NULL);
1da177e4
LT		 950 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
951 if (*mp == ifa->ifa_mask &&
952 inet_ifa_match(rt->rt_src, ifa))
953 break;
954 }
955 if (!ifa && net_ratelimit()) {
a7d632b6
YH		 956 printk(KERN_INFO "Wrong address mask " NIPQUAD_FMT " from "
957 "%s/" NIPQUAD_FMT "\n",
1da177e4
LT		 958 NIPQUAD(*mp), dev->name, NIPQUAD(rt->rt_src));
959 }
960 }
961 rcu_read_unlock();
962 in_dev_put(in_dev);
963out:;
964}
965
966static void icmp_discard(struct sk_buff *skb)
967{
968}
969
970/*
971 * Deal with incoming ICMP packets.
972 */
973int icmp_rcv(struct sk_buff *skb)
974{
975 struct icmphdr *icmph;
ee6b9673 976 struct rtable *rt = skb->rtable;
1da177e4 977
aebcf82c 978 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
8b7817f3
HX		 979 int nh;
980
aebcf82c
HX		 981 if (!(skb->sp && skb->sp->xvec[skb->sp->len - 1]->props.flags &
982 XFRM_STATE_ICMP))
983 goto drop;
984
8b7817f3
HX		 985 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
986 goto drop;
987
988 nh = skb_network_offset(skb);
989 skb_set_network_header(skb, sizeof(*icmph));
990
991 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
992 goto drop;
993
994 skb_set_network_header(skb, nh);
995 }
996
1da177e4
LT		 997 ICMP_INC_STATS_BH(ICMP_MIB_INMSGS);
998
999 switch (skb->ip_summed) {
84fa7933 1000 case CHECKSUM_COMPLETE:
d3bc23e7 1001 if (!csum_fold(skb->csum))
1da177e4 1002 break;
fb286bb2 1003 /* fall through */
1da177e4 1004 case CHECKSUM_NONE:
fb286bb2
HX		 1005 skb->csum = 0;
1006 if (__skb_checksum_complete(skb))
1da177e4 1007 goto error;
1da177e4
LT		 1008 }
1009
8cf22943
HX		 1010 if (!pskb_pull(skb, sizeof(*icmph)))
1011 goto error;
1da177e4 1012
88c7664f 1013 icmph = icmp_hdr(skb);
1da177e4 1014
96793b48 1015 ICMPMSGIN_INC_STATS_BH(icmph->type);
1da177e4
LT		 1016 /*
1017 * 18 is the highest 'known' ICMP type. Anything else is a mystery
1018 *
1019 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
1020 * discarded.
1021 */
1022 if (icmph->type > NR_ICMP_TYPES)
1023 goto error;
1024
1025
1026 /*
1027 * Parse the ICMP message
1028 */
1029
e905a9ed 1030 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
b34a95ee
PE		 1031 struct net *net;
1032
789e41e6 1033 net = dev_net(rt->u.dst.dev);
1da177e4
LT		 1034 /*
1035 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
1036 * silently ignored (we let user decide with a sysctl).
1037 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1038 * discarded if to broadcast/multicast.
1039 */
4c866aa7
AK		 1040 if ((icmph->type == ICMP_ECHO ||
1041 icmph->type == ICMP_TIMESTAMP) &&
b34a95ee 1042 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
1da177e4
LT		 1043 goto error;
1044 }
1045 if (icmph->type != ICMP_ECHO &&
1046 icmph->type != ICMP_TIMESTAMP &&
1047 icmph->type != ICMP_ADDRESS &&
1048 icmph->type != ICMP_ADDRESSREPLY) {
1049 goto error;
e905a9ed 1050 }
1da177e4
LT		 1051 }
1052
1da177e4
LT		 1053 icmp_pointers[icmph->type].handler(skb);
1054
1055drop:
1056 kfree_skb(skb);
1057 return 0;
1058error:
1059 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
1060 goto drop;
1061}
1062
1063/*
1064 * This table is the definition of how we handle ICMP.
1065 */
9b5b5cff 1066static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
1da177e4 1067 [ICMP_ECHOREPLY] = {
1da177e4
LT		 1068 .handler = icmp_discard,
1069 },
1070 [1] = {
1da177e4
LT		 1071 .handler = icmp_discard,
1072 .error = 1,
1073 },
1074 [2] = {
1da177e4
LT		 1075 .handler = icmp_discard,
1076 .error = 1,
1077 },
1078 [ICMP_DEST_UNREACH] = {
1da177e4
LT		 1079 .handler = icmp_unreach,
1080 .error = 1,
1081 },
1082 [ICMP_SOURCE_QUENCH] = {
1da177e4
LT		 1083 .handler = icmp_unreach,
1084 .error = 1,
1085 },
1086 [ICMP_REDIRECT] = {
1da177e4
LT		 1087 .handler = icmp_redirect,
1088 .error = 1,
1089 },
1090 [6] = {
1da177e4
LT		 1091 .handler = icmp_discard,
1092 .error = 1,
1093 },
1094 [7] = {
1da177e4
LT		 1095 .handler = icmp_discard,
1096 .error = 1,
1097 },
1098 [ICMP_ECHO] = {
1da177e4
LT		 1099 .handler = icmp_echo,
1100 },
1101 [9] = {
1da177e4
LT		 1102 .handler = icmp_discard,
1103 .error = 1,
1104 },
1105 [10] = {
1da177e4
LT		 1106 .handler = icmp_discard,
1107 .error = 1,
1108 },
1109 [ICMP_TIME_EXCEEDED] = {
1da177e4
LT		 1110 .handler = icmp_unreach,
1111 .error = 1,
1112 },
1113 [ICMP_PARAMETERPROB] = {
1da177e4
LT		 1114 .handler = icmp_unreach,
1115 .error = 1,
1116 },
1117 [ICMP_TIMESTAMP] = {
1da177e4
LT		 1118 .handler = icmp_timestamp,
1119 },
1120 [ICMP_TIMESTAMPREPLY] = {
1da177e4
LT		 1121 .handler = icmp_discard,
1122 },
1123 [ICMP_INFO_REQUEST] = {
1da177e4
LT		 1124 .handler = icmp_discard,
1125 },
e905a9ed 1126 [ICMP_INFO_REPLY] = {
1da177e4
LT		 1127 .handler = icmp_discard,
1128 },
1129 [ICMP_ADDRESS] = {
1da177e4
LT		 1130 .handler = icmp_address,
1131 },
1132 [ICMP_ADDRESSREPLY] = {
1da177e4
LT		 1133 .handler = icmp_address_reply,
1134 },
1135};
1136
4a6ad7a1 1137static void __net_exit icmp_sk_exit(struct net *net)
1da177e4 1138{
1da177e4
LT		 1139 int i;
1140
5c8cafd6 1141 for_each_possible_cpu(i)
c1e9894d 1142 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
4a6ad7a1
DL		 1143 kfree(net->ipv4.icmp_sk);
1144 net->ipv4.icmp_sk = NULL;
a5710d65 1145}
1da177e4 1146
4a6ad7a1 1147int __net_init icmp_sk_init(struct net *net)
a5710d65 1148{
a5710d65
DL		 1149 int i, err;
1150
4a6ad7a1
DL		 1151 net->ipv4.icmp_sk =
1152 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1153 if (net->ipv4.icmp_sk == NULL)
79c91159
DL		 1154 return -ENOMEM;
1155
a5710d65 1156 for_each_possible_cpu(i) {
1e3cf683 1157 struct sock *sk;
1da177e4 1158
c1e9894d
DL		 1159 err = inet_ctl_sock_create(&sk, PF_INET,
1160 SOCK_RAW, IPPROTO_ICMP, net);
1da177e4 1161 if (err < 0)
a5710d65 1162 goto fail;
1da177e4 1163
c1e9894d 1164 net->ipv4.icmp_sk[i] = sk;
1da177e4
LT		 1165
1166 /* Enough space for 2 64K ICMP packets, including
1167 * sk_buff struct overhead.
1168 */
1e3cf683 1169 sk->sk_sndbuf =
1da177e4
LT		 1170 (2 * ((64 * 1024) + sizeof(struct sk_buff)));
1171
c1e9894d 1172 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
1da177e4 1173 }
a24022e1
PE		 1174
1175 /* Control parameters for ECHO replies. */
1176 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1177 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1178
1179 /* Control parameter - ignore bogus broadcast responses? */
1180 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1181
1182 /*
1183 * Configurable global rate limit.
1184 *
1185 * ratelimit defines tokens/packet consumed for dst->rate_token
1186 * bucket ratemask defines which icmp types are ratelimited by
1187 * setting it's bit position.
1188 *
1189 * default:
1190 * dest unreachable (3), source quench (4),
1191 * time exceeded (11), parameter problem (12)
1192 */
1193
1194 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1195 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1196 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1197
a5710d65
DL		 1198 return 0;
1199
1200fail:
1d1c8d13 1201 for_each_possible_cpu(i)
c1e9894d 1202 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
1d1c8d13 1203 kfree(net->ipv4.icmp_sk);
a5710d65 1204 return err;
1da177e4
LT		 1205}
1206
4a6ad7a1
DL		 1207static struct pernet_operations __net_initdata icmp_sk_ops = {
1208 .init = icmp_sk_init,
1209 .exit = icmp_sk_exit,
1210};
1211
1212int __init icmp_init(void)
1213{
1214 return register_pernet_device(&icmp_sk_ops);
1215}
1216
1da177e4
LT		 1217EXPORT_SYMBOL(icmp_err_convert);
1218EXPORT_SYMBOL(icmp_send);
1219EXPORT_SYMBOL(icmp_statistics);
1220EXPORT_SYMBOL(xrlim_allow);
kernel source for telekom puls (alcatel/mediatek)