blob: 5bc13fe816d125780d1b5518560dda663817e80e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET3: Implementation of the ICMP protocol layer.
3 *
Alan Cox113aa832008-10-13 19:01:08 -07004 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07006 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 * Some of the function names and the icmp unreach table for this
12 * module were derived from [icmp.c 1.0.11 06/02/93] by
13 * Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
14 * Other than that this module is a complete rewrite.
15 *
16 * Fixes:
17 * Clemens Fruhwirth : introduce global icmp rate limiting
18 * with icmp type masking ability instead
19 * of broken per type icmp timeouts.
20 * Mike Shaver : RFC1122 checks.
21 * Alan Cox : Multicast ping reply as self.
22 * Alan Cox : Fix atomicity lockup in ip_build_xmit
23 * call.
24 * Alan Cox : Added 216,128 byte paths to the MTU
25 * code.
26 * Martin Mares : RFC1812 checks.
27 * Martin Mares : Can be configured to follow redirects
28 * if acting as a router _without_ a
29 * routing protocol (RFC 1812).
30 * Martin Mares : Echo requests may be configured to
31 * be ignored (RFC 1812).
32 * Martin Mares : Limitation of ICMP error message
33 * transmit rate (RFC 1812).
34 * Martin Mares : TOS and Precedence set correctly
35 * (RFC 1812).
36 * Martin Mares : Now copying as much data from the
37 * original packet as we can without
38 * exceeding 576 bytes (RFC 1812).
39 * Willy Konynenberg : Transparent proxying support.
40 * Keith Owens : RFC1191 correction for 4.2BSD based
41 * path MTU bug.
42 * Thomas Quinot : ICMP Dest Unreach codes up to 15 are
43 * valid (RFC 1812).
44 * Andi Kleen : Check all packet lengths properly
45 * and moved all kfree_skb() up to
46 * icmp_rcv.
47 * Andi Kleen : Move the rate limit bookkeeping
48 * into the dest entry and use a token
49 * bucket filter (thanks to ANK). Make
50 * the rates sysctl configurable.
51 * Yu Tianli : Fixed two ugly bugs in icmp_send
52 * - IP option length was accounted wrongly
53 * - ICMP header length was not accounted
54 * at all.
55 * Tristan Greaves : Added sysctl option to ignore bogus
56 * broadcast responses from broken routers.
57 *
58 * To Fix:
59 *
60 * - Should use skb_pull() instead of all the manual checking.
61 * This would also greatly simply some upper layer error handlers. --AK
62 *
63 */
64
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#include <linux/module.h>
66#include <linux/types.h>
67#include <linux/jiffies.h>
68#include <linux/kernel.h>
69#include <linux/fcntl.h>
70#include <linux/socket.h>
71#include <linux/in.h>
72#include <linux/inet.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020073#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070074#include <linux/netdevice.h>
75#include <linux/string.h>
76#include <linux/netfilter_ipv4.h>
77#include <net/snmp.h>
78#include <net/ip.h>
79#include <net/route.h>
80#include <net/protocol.h>
81#include <net/icmp.h>
82#include <net/tcp.h>
83#include <net/udp.h>
84#include <net/raw.h>
85#include <linux/skbuff.h>
86#include <net/sock.h>
87#include <linux/errno.h>
88#include <linux/timer.h>
89#include <linux/init.h>
90#include <asm/system.h>
91#include <asm/uaccess.h>
92#include <net/checksum.h>
Herbert Xu8b7817f2007-12-12 10:44:43 -080093#include <net/xfrm.h>
Denis V. Lunevc1e98942008-04-03 14:29:00 -070094#include <net/inet_common.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96/*
97 * Build xmit assembly blocks
98 */
99
100struct icmp_bxm {
101 struct sk_buff *skb;
102 int offset;
103 int data_len;
104
105 struct {
106 struct icmphdr icmph;
Al Virob03d73e2006-09-27 18:38:30 -0700107 __be32 times[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 } data;
109 int head_len;
110 struct ip_options replyopts;
111 unsigned char optbuf[40];
112};
113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114/* An array of errno for error messages from dest unreach. */
115/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
116
117struct icmp_err icmp_err_convert[] = {
118 {
119 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
120 .fatal = 0,
121 },
122 {
123 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
124 .fatal = 0,
125 },
126 {
127 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
128 .fatal = 1,
129 },
130 {
131 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
132 .fatal = 1,
133 },
134 {
135 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
136 .fatal = 0,
137 },
138 {
139 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
140 .fatal = 0,
141 },
142 {
143 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
144 .fatal = 1,
145 },
146 {
147 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
148 .fatal = 1,
149 },
150 {
151 .errno = ENONET, /* ICMP_HOST_ISOLATED */
152 .fatal = 1,
153 },
154 {
155 .errno = ENETUNREACH, /* ICMP_NET_ANO */
156 .fatal = 1,
157 },
158 {
159 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
160 .fatal = 1,
161 },
162 {
163 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
164 .fatal = 0,
165 },
166 {
167 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
168 .fatal = 0,
169 },
170 {
171 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
172 .fatal = 1,
173 },
174 {
175 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
176 .fatal = 1,
177 },
178 {
179 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
180 .fatal = 1,
181 },
182};
183
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184/*
185 * ICMP control array. This specifies what to do with each ICMP.
186 */
187
188struct icmp_control {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 void (*handler)(struct sk_buff *skb);
190 short error; /* This ICMP is classed as an error message */
191};
192
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800193static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
195/*
196 * The ICMP socket(s). This is the most convenient way to flow control
197 * our ICMP output as well as maintain a clean interface throughout
198 * all layers. All Socketless IP sends will soon be gone.
199 *
200 * On SMP we have one ICMP socket per-cpu.
201 */
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800202static struct sock *icmp_sk(struct net *net)
203{
204 return net->ipv4.icmp_sk[smp_processor_id()];
205}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700207static inline struct sock *icmp_xmit_lock(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208{
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700209 struct sock *sk;
210
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 local_bh_disable();
212
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700213 sk = icmp_sk(net);
214
Denis V. Lunev405666d2008-02-29 11:16:46 -0800215 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 /* This can happen if the output path signals a
217 * dst_link_failure() for an outgoing ICMP packet.
218 */
219 local_bh_enable();
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700220 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 }
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700222 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
224
Denis V. Lunev405666d2008-02-29 11:16:46 -0800225static inline void icmp_xmit_unlock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226{
Denis V. Lunev405666d2008-02-29 11:16:46 -0800227 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228}
229
230/*
231 * Send an ICMP frame.
232 */
233
234/*
235 * Check transmit rate limitation for given message.
236 * The rate information is held in the destination cache now.
237 * This function is generic and could be used for other purposes
238 * too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
239 *
240 * Note that the same dst_entry fields are modified by functions in
241 * route.c too, but these work for packet destinations while xrlim_allow
242 * works for icmp destinations. This means the rate limiting information
243 * for one "ip object" is shared - and these ICMPs are twice limited:
244 * by source and by destination.
245 *
246 * RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
247 * SHOULD allow setting of rate limits
248 *
249 * Shared between ICMPv4 and ICMPv6.
250 */
251#define XRLIM_BURST_FACTOR 6
252int xrlim_allow(struct dst_entry *dst, int timeout)
253{
Eric Dumazet69a73822008-01-22 06:18:34 -0800254 unsigned long now, token = dst->rate_tokens;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 int rc = 0;
256
257 now = jiffies;
Eric Dumazet69a73822008-01-22 06:18:34 -0800258 token += now - dst->rate_last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 dst->rate_last = now;
Eric Dumazet69a73822008-01-22 06:18:34 -0800260 if (token > XRLIM_BURST_FACTOR * timeout)
261 token = XRLIM_BURST_FACTOR * timeout;
262 if (token >= timeout) {
263 token -= timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 rc = 1;
265 }
Eric Dumazet69a73822008-01-22 06:18:34 -0800266 dst->rate_tokens = token;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 return rc;
268}
269
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700270static inline int icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
271 int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272{
273 struct dst_entry *dst = &rt->u.dst;
274 int rc = 1;
275
276 if (type > NR_ICMP_TYPES)
277 goto out;
278
279 /* Don't limit PMTU discovery. */
280 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
281 goto out;
282
283 /* No rate limit on loopback */
284 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900285 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
287 /* Limit if icmp type is enabled in ratemask. */
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700288 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask)
289 rc = xrlim_allow(dst, net->ipv4.sysctl_icmp_ratelimit);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290out:
291 return rc;
292}
293
294/*
295 * Maintain the counters used in the SNMP statistics for outgoing ICMP
296 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700297void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700299 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700300 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301}
302
303/*
304 * Checksum each fragment, and on the first include the headers and final
305 * checksum.
306 */
307static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
308 struct sk_buff *skb)
309{
310 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800311 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
313 csum = skb_copy_and_csum_bits(icmp_param->skb,
314 icmp_param->offset + offset,
315 to, len, 0);
316
317 skb->csum = csum_block_add(skb->csum, csum, odd);
318 if (icmp_pointers[icmp_param->data.icmph.type].error)
319 nf_ct_attach(skb, icmp_param->skb);
320 return 0;
321}
322
323static void icmp_push_reply(struct icmp_bxm *icmp_param,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800324 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800326 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 struct sk_buff *skb;
328
Eric Dumazet2e77d892008-11-24 15:52:46 -0800329 sk = icmp_sk(dev_net((*rt)->u.dst.dev));
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800330 if (ip_append_data(sk, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900331 icmp_param->data_len+icmp_param->head_len,
332 icmp_param->head_len,
333 ipc, rt, MSG_DONTWAIT) < 0)
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800334 ip_flush_pending_frames(sk);
335 else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300336 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800337 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 struct sk_buff *skb1;
339
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800340 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 csum = csum_add(csum, skb1->csum);
342 }
343 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
344 (char *)icmph,
345 icmp_param->head_len, csum);
346 icmph->checksum = csum_fold(csum);
347 skb->ip_summed = CHECKSUM_NONE;
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800348 ip_push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 }
350}
351
352/*
353 * Driving logic for building and sending ICMP messages.
354 */
355
356static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
357{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000359 struct rtable *rt = skb_rtable(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900360 struct net *net = dev_net(rt->u.dst.dev);
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700361 struct sock *sk;
362 struct inet_sock *inet;
Al Viro3ca3c682006-09-27 18:28:07 -0700363 __be32 daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
365 if (ip_options_echo(&icmp_param->replyopts, skb))
Hormsf00c4012006-02-02 17:03:18 -0800366 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700368 sk = icmp_xmit_lock(net);
369 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700371 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
373 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700375 inet->tos = ip_hdr(skb)->tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 daddr = ipc.addr = rt->rt_src;
377 ipc.opt = NULL;
Patrick Ohly51f31ca2009-02-12 05:03:39 +0000378 ipc.shtx.flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 if (icmp_param->replyopts.optlen) {
380 ipc.opt = &icmp_param->replyopts;
381 if (ipc.opt->srr)
382 daddr = icmp_param->replyopts.faddr;
383 }
384 {
385 struct flowi fl = { .nl_u = { .ip4_u =
386 { .daddr = daddr,
387 .saddr = rt->rt_spec_dst,
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700388 .tos = RT_TOS(ip_hdr(skb)->tos) } },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 .proto = IPPROTO_ICMP };
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -0700390 security_skb_classify_flow(skb, &fl);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800391 if (ip_route_output_key(net, &rt, &fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 goto out_unlock;
393 }
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700394 if (icmpv4_xrlim_allow(net, rt, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 icmp_param->data.icmph.code))
Eric Dumazet2e77d892008-11-24 15:52:46 -0800396 icmp_push_reply(icmp_param, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 ip_rt_put(rt);
398out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800399 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400}
401
402
403/*
404 * Send an ICMP message in response to a situation
405 *
406 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
407 * MAY send more (we do).
408 * MUST NOT change this header information.
409 * MUST NOT reply to a multicast/broadcast IP address.
410 * MUST NOT reply to a multicast/broadcast MAC address.
411 * MUST reply to only the first fragment.
412 */
413
Al Viroe4883012006-09-26 22:17:28 -0700414void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415{
416 struct iphdr *iph;
417 int room;
418 struct icmp_bxm icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000419 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 struct ipcm_cookie ipc;
Al Viroa61ced52006-09-26 21:27:54 -0700421 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 u8 tos;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800423 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800424 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
426 if (!rt)
427 goto out;
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900428 net = dev_net(rt->u.dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429
430 /*
431 * Find the original header. It is expected to be valid, of course.
432 * Check this, icmp_send is called from the most obscure devices
433 * sometimes.
434 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700435 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700437 if ((u8 *)iph < skb_in->head ||
438 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 goto out;
440
441 /*
442 * No replies to physical multicast/broadcast
443 */
444 if (skb_in->pkt_type != PACKET_HOST)
445 goto out;
446
447 /*
448 * Now check at the protocol level
449 */
450 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
451 goto out;
452
453 /*
454 * Only reply to fragment 0. We byte re-order the constant
455 * mask for efficiency.
456 */
457 if (iph->frag_off & htons(IP_OFFSET))
458 goto out;
459
460 /*
461 * If we send an ICMP error to an ICMP error a mess would result..
462 */
463 if (icmp_pointers[type].error) {
464 /*
465 * We are an error, check if we are replying to an
466 * ICMP error
467 */
468 if (iph->protocol == IPPROTO_ICMP) {
469 u8 _inner_type, *itp;
470
471 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700472 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 (iph->ihl << 2) +
474 offsetof(struct icmphdr,
475 type) -
476 skb_in->data,
477 sizeof(_inner_type),
478 &_inner_type);
479 if (itp == NULL)
480 goto out;
481
482 /*
483 * Assume any unknown ICMP type is an error. This
484 * isn't specified by the RFC, but think about it..
485 */
486 if (*itp > NR_ICMP_TYPES ||
487 icmp_pointers[*itp].error)
488 goto out;
489 }
490 }
491
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700492 sk = icmp_xmit_lock(net);
493 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 return;
495
496 /*
497 * Construct source address and options.
498 */
499
500 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700501 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700502 struct net_device *dev = NULL;
503
Pavel Emelyanova24022e2008-03-26 01:55:37 -0700504 if (rt->fl.iif &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700505 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800506 dev = dev_get_by_index(net, rt->fl.iif);
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700507
508 if (dev) {
509 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
510 dev_put(dev);
511 } else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700512 saddr = 0;
513 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
515 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
516 IPTOS_PREC_INTERNETCONTROL) :
517 iph->tos;
518
519 if (ip_options_echo(&icmp_param.replyopts, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800520 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
522
523 /*
524 * Prepare data for ICMP header.
525 */
526
527 icmp_param.data.icmph.type = type;
528 icmp_param.data.icmph.code = code;
529 icmp_param.data.icmph.un.gateway = info;
530 icmp_param.data.icmph.checksum = 0;
531 icmp_param.skb = skb_in;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700532 icmp_param.offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800533 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 ipc.addr = iph->saddr;
535 ipc.opt = &icmp_param.replyopts;
Patrick Ohly51f31ca2009-02-12 05:03:39 +0000536 ipc.shtx.flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
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 };
Herbert Xu8b7817f2007-12-12 10:44:43 -0800557 int err;
558 struct rtable *rt2;
559
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -0700560 security_skb_classify_flow(skb_in, &fl);
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800561 if (__ip_route_output_key(net, &rt, &fl))
Herbert Xu8b7817f2007-12-12 10:44:43 -0800562 goto out_unlock;
563
564 /* No need to clone since we're just using its address. */
565 rt2 = rt;
566
Alexey Dobriyan52479b62008-11-25 17:35:18 -0800567 err = xfrm_lookup(net, (struct dst_entry **)&rt, &fl, NULL, 0);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800568 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))
Herbert Xuaf268182008-04-03 12:52:19 -0700581 goto relookup_failed;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800582
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800583 if (inet_addr_type(net, fl.fl4_src) == RTN_LOCAL)
584 err = __ip_route_output_key(net, &rt2, &fl);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800585 else {
586 struct flowi fl2 = {};
587 struct dst_entry *odst;
588
589 fl2.fl4_dst = fl.fl4_src;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800590 if (ip_route_output_key(net, &rt2, &fl2))
Herbert Xuaf268182008-04-03 12:52:19 -0700591 goto relookup_failed;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800592
593 /* Ugh! */
Eric Dumazetadf30902009-06-02 05:19:30 +0000594 odst = skb_dst(skb_in);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800595 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);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000599 rt2 = skb_rtable(skb_in);
Eric Dumazetadf30902009-06-02 05:19:30 +0000600 skb_dst_set(skb_in, odst);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800601 }
602
603 if (err)
Herbert Xuaf268182008-04-03 12:52:19 -0700604 goto relookup_failed;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800605
Alexey Dobriyan52479b62008-11-25 17:35:18 -0800606 err = xfrm_lookup(net, (struct dst_entry **)&rt2, &fl, NULL,
Herbert Xu8b7817f2007-12-12 10:44:43 -0800607 XFRM_LOOKUP_ICMP);
Herbert Xuaf268182008-04-03 12:52:19 -0700608 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:
Herbert Xu8b7817f2007-12-12 10:44:43 -0800617 if (!rt)
618 goto out_unlock;
Herbert Xuaf268182008-04-03 12:52:19 -0700619 break;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800620 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 }
622
Herbert Xu8b7817f2007-12-12 10:44:43 -0800623route_done:
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700624 if (!icmpv4_xrlim_allow(net, rt, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 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
Eric Dumazet2e77d892008-11-24 15:52:46 -0800640 icmp_push_reply(&icmp_param, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641ende:
642 ip_rt_put(rt);
643out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800644 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645out:;
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;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000658 const struct net_protocol *ipprot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800660 struct net *net;
661
Eric Dumazetadf30902009-06-02 05:19:30 +0000662 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663
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
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300673 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 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) {
Harvey Harrison673d57e2008-10-31 00:53:57 -0700688 LIMIT_NETDEBUG(KERN_INFO "ICMP: %pI4: fragmentation needed and DF set.\n",
689 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 } else {
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800691 info = ip_rt_frag_needed(net, iph,
Timo Teras0010e462008-04-29 03:32:25 -0700692 ntohs(icmph->un.frag.mtu),
693 skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 if (!info)
695 goto out;
696 }
697 break;
698 case ICMP_SR_FAILED:
Harvey Harrison673d57e2008-10-31 00:53:57 -0700699 LIMIT_NETDEBUG(KERN_INFO "ICMP: %pI4: Source Route Failed.\n",
700 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 break;
702 default:
703 break;
704 }
705 if (icmph->code > NR_ICMP_UNREACH)
706 goto out;
707 } else if (icmph->type == ICMP_PARAMETERPROB)
708 info = ntohl(icmph->un.gateway) >> 24;
709
710 /*
711 * Throw it at our lower layers
712 *
713 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
714 * header.
715 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
716 * transport layer.
717 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
718 * transport layer.
719 */
720
721 /*
722 * Check the other end isnt violating RFC 1122. Some routers send
723 * bogus responses to broadcast frames. If you see this message
724 * first check your netmask matches at both ends, if it does then
725 * get the other vendor to fix their kit.
726 */
727
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700728 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800729 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 if (net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700731 printk(KERN_WARNING "%pI4 sent an invalid ICMP "
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 "type %u, code %u "
Harvey Harrison673d57e2008-10-31 00:53:57 -0700733 "error to a broadcast: %pI4 on %s\n",
734 &ip_hdr(skb)->saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 icmph->type, icmph->code,
Harvey Harrison673d57e2008-10-31 00:53:57 -0700736 &iph->daddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 skb->dev->name);
738 goto out;
739 }
740
741 /* Checkin full IP header plus 8 bytes of protocol to
742 * avoid additional coding at protocol handlers.
743 */
744 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
745 goto out;
746
747 iph = (struct iphdr *)skb->data;
748 protocol = iph->protocol;
749
750 /*
751 * Deliver ICMP message to raw sockets. Pretty useless feature?
752 */
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800753 raw_icmp_error(skb, protocol, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 hash = protocol & (MAX_INET_PROTOS - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 rcu_read_lock();
757 ipprot = rcu_dereference(inet_protos[hash]);
758 if (ipprot && ipprot->err_handler)
759 ipprot->err_handler(skb, info);
760 rcu_read_unlock();
761
762out:
763 return;
764out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700765 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 goto out;
767}
768
769
770/*
771 * Handle ICMP_REDIRECT.
772 */
773
774static void icmp_redirect(struct sk_buff *skb)
775{
776 struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777
778 if (skb->len < sizeof(struct iphdr))
779 goto out_err;
780
781 /*
782 * Get the copied header of the packet that caused the redirect
783 */
784 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
785 goto out;
786
787 iph = (struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300789 switch (icmp_hdr(skb)->code & 7) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 case ICMP_REDIR_NET:
791 case ICMP_REDIR_NETTOS:
792 /*
793 * As per RFC recommendations now handle it as a host redirect.
794 */
795 case ICMP_REDIR_HOST:
796 case ICMP_REDIR_HOSTTOS:
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700797 ip_rt_redirect(ip_hdr(skb)->saddr, iph->daddr,
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300798 icmp_hdr(skb)->un.gateway,
Ilia Sotnikovcef26852006-03-25 01:38:55 -0800799 iph->saddr, skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900801 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802out:
803 return;
804out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700805 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 goto out;
807}
808
809/*
810 * Handle ICMP_ECHO ("ping") requests.
811 *
812 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
813 * requests.
814 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
815 * included in the reply.
816 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
817 * echo requests, MUST have default=NOT.
818 * See also WRT handling of options once they are done and working.
819 */
820
821static void icmp_echo(struct sk_buff *skb)
822{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700823 struct net *net;
824
Eric Dumazetadf30902009-06-02 05:19:30 +0000825 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700826 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 struct icmp_bxm icmp_param;
828
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300829 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
831 icmp_param.skb = skb;
832 icmp_param.offset = 0;
833 icmp_param.data_len = skb->len;
834 icmp_param.head_len = sizeof(struct icmphdr);
835 icmp_reply(&icmp_param, skb);
836 }
837}
838
839/*
840 * Handle ICMP Timestamp requests.
841 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
842 * SHOULD be in the kernel for minimum random latency.
843 * MUST be accurate to a few minutes.
844 * MUST be updated at least at 15Hz.
845 */
846static void icmp_timestamp(struct sk_buff *skb)
847{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700848 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 struct icmp_bxm icmp_param;
850 /*
851 * Too short.
852 */
853 if (skb->len < 4)
854 goto out_err;
855
856 /*
857 * Fill in the current time as ms since midnight UT:
858 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700859 getnstimeofday(&tv);
860 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
861 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 icmp_param.data.times[2] = icmp_param.data.times[1];
863 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
864 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300865 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
867 icmp_param.data.icmph.code = 0;
868 icmp_param.skb = skb;
869 icmp_param.offset = 0;
870 icmp_param.data_len = 0;
871 icmp_param.head_len = sizeof(struct icmphdr) + 12;
872 icmp_reply(&icmp_param, skb);
873out:
874 return;
875out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000876 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 goto out;
878}
879
880
881/*
882 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
883 *
884 * RFC1122 (3.2.2.9). A host MUST only send replies to
885 * ADDRESS_MASK requests if it's been configured as an address mask
886 * agent. Receiving a request doesn't constitute implicit permission to
887 * act as one. Of course, implementing this correctly requires (SHOULD)
888 * a way to turn the functionality on and off. Another one for sysctl(),
889 * I guess. -- MS
890 *
891 * RFC1812 (4.3.3.9). A router MUST implement it.
892 * A router SHOULD have switch turning it on/off.
893 * This switch MUST be ON by default.
894 *
895 * Gratuitous replies, zero-source replies are not implemented,
896 * that complies with RFC. DO NOT implement them!!! All the idea
897 * of broadcast addrmask replies as specified in RFC950 is broken.
898 * The problem is that it is not uncommon to have several prefixes
899 * on one physical interface. Moreover, addrmask agent can even be
900 * not aware of existing another prefixes.
901 * If source is zero, addrmask agent cannot choose correct prefix.
902 * Gratuitous mask announcements suffer from the same problem.
903 * RFC1812 explains it, but still allows to use ADDRMASK,
904 * that is pretty silly. --ANK
905 *
906 * All these rules are so bizarre, that I removed kernel addrmask
907 * support at all. It is wrong, it is obsolete, nobody uses it in
908 * any case. --ANK
909 *
910 * Furthermore you can do it with a usermode address agent program
911 * anyway...
912 */
913
914static void icmp_address(struct sk_buff *skb)
915{
916#if 0
917 if (net_ratelimit())
918 printk(KERN_DEBUG "a guy asks for address mask. Who is it?\n");
919#endif
920}
921
922/*
923 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
924 * loudly if an inconsistency is found.
925 */
926
927static void icmp_address_reply(struct sk_buff *skb)
928{
Eric Dumazet511c3f92009-06-02 05:14:27 +0000929 struct rtable *rt = skb_rtable(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 struct net_device *dev = skb->dev;
931 struct in_device *in_dev;
932 struct in_ifaddr *ifa;
933
934 if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC))
935 goto out;
936
937 in_dev = in_dev_get(dev);
938 if (!in_dev)
939 goto out;
940 rcu_read_lock();
941 if (in_dev->ifa_list &&
942 IN_DEV_LOG_MARTIANS(in_dev) &&
943 IN_DEV_FORWARD(in_dev)) {
Al Viroa144ea42006-09-28 18:00:55 -0700944 __be32 _mask, *mp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
946 mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask);
Kris Katterjohn09a62662006-01-08 22:24:28 -0800947 BUG_ON(mp == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
949 if (*mp == ifa->ifa_mask &&
950 inet_ifa_match(rt->rt_src, ifa))
951 break;
952 }
953 if (!ifa && net_ratelimit()) {
Harvey Harrison673d57e2008-10-31 00:53:57 -0700954 printk(KERN_INFO "Wrong address mask %pI4 from %s/%pI4\n",
955 mp, dev->name, &rt->rt_src);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 }
957 }
958 rcu_read_unlock();
959 in_dev_put(in_dev);
960out:;
961}
962
963static void icmp_discard(struct sk_buff *skb)
964{
965}
966
967/*
968 * Deal with incoming ICMP packets.
969 */
970int icmp_rcv(struct sk_buff *skb)
971{
972 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000973 struct rtable *rt = skb_rtable(skb);
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700974 struct net *net = dev_net(rt->u.dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
Herbert Xuaebcf822007-12-12 18:54:16 -0800976 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700977 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800978 int nh;
979
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700980 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800981 XFRM_STATE_ICMP))
982 goto drop;
983
Herbert Xu8b7817f2007-12-12 10:44:43 -0800984 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
985 goto drop;
986
987 nh = skb_network_offset(skb);
988 skb_set_network_header(skb, sizeof(*icmph));
989
990 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
991 goto drop;
992
993 skb_set_network_header(skb, nh);
994 }
995
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700996 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997
998 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700999 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -08001000 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 break;
Herbert Xufb286bb2005-11-10 13:01:24 -08001002 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -08001004 skb->csum = 0;
1005 if (__skb_checksum_complete(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 }
1008
Herbert Xu8cf22942008-02-05 03:15:50 -08001009 if (!pskb_pull(skb, sizeof(*icmph)))
1010 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -03001012 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013
Pavel Emelyanovf66ac032008-07-14 23:04:00 -07001014 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 /*
1016 * 18 is the highest 'known' ICMP type. Anything else is a mystery
1017 *
1018 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
1019 * discarded.
1020 */
1021 if (icmph->type > NR_ICMP_TYPES)
1022 goto error;
1023
1024
1025 /*
1026 * Parse the ICMP message
1027 */
1028
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001029 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 /*
1031 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
1032 * silently ignored (we let user decide with a sysctl).
1033 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1034 * discarded if to broadcast/multicast.
1035 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -07001036 if ((icmph->type == ICMP_ECHO ||
1037 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -07001038 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 goto error;
1040 }
1041 if (icmph->type != ICMP_ECHO &&
1042 icmph->type != ICMP_TIMESTAMP &&
1043 icmph->type != ICMP_ADDRESS &&
1044 icmph->type != ICMP_ADDRESSREPLY) {
1045 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001046 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 }
1048
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 icmp_pointers[icmph->type].handler(skb);
1050
1051drop:
1052 kfree_skb(skb);
1053 return 0;
1054error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001055 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 goto drop;
1057}
1058
1059/*
1060 * This table is the definition of how we handle ICMP.
1061 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -08001062static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 [ICMP_ECHOREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 .handler = icmp_discard,
1065 },
1066 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 .handler = icmp_discard,
1068 .error = 1,
1069 },
1070 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 .handler = icmp_discard,
1072 .error = 1,
1073 },
1074 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 .handler = icmp_unreach,
1076 .error = 1,
1077 },
1078 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 .handler = icmp_unreach,
1080 .error = 1,
1081 },
1082 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 .handler = icmp_redirect,
1084 .error = 1,
1085 },
1086 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 .handler = icmp_discard,
1088 .error = 1,
1089 },
1090 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 .handler = icmp_discard,
1092 .error = 1,
1093 },
1094 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 .handler = icmp_echo,
1096 },
1097 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 .handler = icmp_discard,
1099 .error = 1,
1100 },
1101 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 .handler = icmp_discard,
1103 .error = 1,
1104 },
1105 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 .handler = icmp_unreach,
1107 .error = 1,
1108 },
1109 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 .handler = icmp_unreach,
1111 .error = 1,
1112 },
1113 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 .handler = icmp_timestamp,
1115 },
1116 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 .handler = icmp_discard,
1118 },
1119 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 .handler = icmp_discard,
1121 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001122 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 .handler = icmp_discard,
1124 },
1125 [ICMP_ADDRESS] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 .handler = icmp_address,
1127 },
1128 [ICMP_ADDRESSREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 .handler = icmp_address_reply,
1130 },
1131};
1132
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001133static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 int i;
1136
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001137 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001138 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001139 kfree(net->ipv4.icmp_sk);
1140 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001141}
1142
Adrian Bunk263173a2008-04-21 02:31:23 -07001143static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001144{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001145 int i, err;
1146
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001147 net->ipv4.icmp_sk =
1148 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1149 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001150 return -ENOMEM;
1151
Denis V. Luneva5710d62008-02-29 11:14:50 -08001152 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001153 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001155 err = inet_ctl_sock_create(&sk, PF_INET,
1156 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001158 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001160 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161
1162 /* Enough space for 2 64K ICMP packets, including
1163 * sk_buff struct overhead.
1164 */
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001165 sk->sk_sndbuf =
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 (2 * ((64 * 1024) + sizeof(struct sk_buff)));
1167
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001168 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001170
1171 /* Control parameters for ECHO replies. */
1172 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1173 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1174
1175 /* Control parameter - ignore bogus broadcast responses? */
1176 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1177
1178 /*
1179 * Configurable global rate limit.
1180 *
1181 * ratelimit defines tokens/packet consumed for dst->rate_token
1182 * bucket ratemask defines which icmp types are ratelimited by
1183 * setting it's bit position.
1184 *
1185 * default:
1186 * dest unreachable (3), source quench (4),
1187 * time exceeded (11), parameter problem (12)
1188 */
1189
1190 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1191 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1192 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1193
Denis V. Luneva5710d62008-02-29 11:14:50 -08001194 return 0;
1195
1196fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001197 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001198 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001199 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001200 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201}
1202
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001203static struct pernet_operations __net_initdata icmp_sk_ops = {
1204 .init = icmp_sk_init,
1205 .exit = icmp_sk_exit,
1206};
1207
1208int __init icmp_init(void)
1209{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001210 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001211}
1212
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213EXPORT_SYMBOL(icmp_err_convert);
1214EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215EXPORT_SYMBOL(xrlim_allow);