blob: a0d847c7cba5c89aec43b340036bf456468f9e9a [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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090077#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070078#include <net/snmp.h>
79#include <net/ip.h>
80#include <net/route.h>
81#include <net/protocol.h>
82#include <net/icmp.h>
83#include <net/tcp.h>
84#include <net/udp.h>
85#include <net/raw.h>
86#include <linux/skbuff.h>
87#include <net/sock.h>
88#include <linux/errno.h>
89#include <linux/timer.h>
90#include <linux/init.h>
91#include <asm/system.h>
92#include <asm/uaccess.h>
93#include <net/checksum.h>
Herbert Xu8b7817f2007-12-12 10:44:43 -080094#include <net/xfrm.h>
Denis V. Lunevc1e98942008-04-03 14:29:00 -070095#include <net/inet_common.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
97/*
98 * Build xmit assembly blocks
99 */
100
101struct icmp_bxm {
102 struct sk_buff *skb;
103 int offset;
104 int data_len;
105
106 struct {
107 struct icmphdr icmph;
Al Virob03d73e2006-09-27 18:38:30 -0700108 __be32 times[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 } data;
110 int head_len;
111 struct ip_options replyopts;
112 unsigned char optbuf[40];
113};
114
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115/* An array of errno for error messages from dest unreach. */
116/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
117
Alexey Dobriyane7548342010-01-22 10:18:25 +0000118const struct icmp_err icmp_err_convert[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 {
120 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
121 .fatal = 0,
122 },
123 {
124 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
125 .fatal = 0,
126 },
127 {
128 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
129 .fatal = 1,
130 },
131 {
132 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
133 .fatal = 1,
134 },
135 {
136 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
137 .fatal = 0,
138 },
139 {
140 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
141 .fatal = 0,
142 },
143 {
144 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
145 .fatal = 1,
146 },
147 {
148 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
149 .fatal = 1,
150 },
151 {
152 .errno = ENONET, /* ICMP_HOST_ISOLATED */
153 .fatal = 1,
154 },
155 {
156 .errno = ENETUNREACH, /* ICMP_NET_ANO */
157 .fatal = 1,
158 },
159 {
160 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
161 .fatal = 1,
162 },
163 {
164 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
165 .fatal = 0,
166 },
167 {
168 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
169 .fatal = 0,
170 },
171 {
172 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
173 .fatal = 1,
174 },
175 {
176 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
177 .fatal = 1,
178 },
179 {
180 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
181 .fatal = 1,
182 },
183};
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000184EXPORT_SYMBOL(icmp_err_convert);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186/*
187 * ICMP control array. This specifies what to do with each ICMP.
188 */
189
190struct icmp_control {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 void (*handler)(struct sk_buff *skb);
192 short error; /* This ICMP is classed as an error message */
193};
194
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800195static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196
197/*
198 * The ICMP socket(s). This is the most convenient way to flow control
199 * our ICMP output as well as maintain a clean interface throughout
200 * all layers. All Socketless IP sends will soon be gone.
201 *
202 * On SMP we have one ICMP socket per-cpu.
203 */
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800204static struct sock *icmp_sk(struct net *net)
205{
206 return net->ipv4.icmp_sk[smp_processor_id()];
207}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700209static inline struct sock *icmp_xmit_lock(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210{
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700211 struct sock *sk;
212
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 local_bh_disable();
214
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700215 sk = icmp_sk(net);
216
Denis V. Lunev405666d2008-02-29 11:16:46 -0800217 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 /* This can happen if the output path signals a
219 * dst_link_failure() for an outgoing ICMP packet.
220 */
221 local_bh_enable();
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700222 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 }
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700224 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225}
226
Denis V. Lunev405666d2008-02-29 11:16:46 -0800227static inline void icmp_xmit_unlock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228{
Denis V. Lunev405666d2008-02-29 11:16:46 -0800229 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230}
231
232/*
233 * Send an ICMP frame.
234 */
235
236/*
237 * Check transmit rate limitation for given message.
238 * The rate information is held in the destination cache now.
239 * This function is generic and could be used for other purposes
240 * too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
241 *
242 * Note that the same dst_entry fields are modified by functions in
243 * route.c too, but these work for packet destinations while xrlim_allow
244 * works for icmp destinations. This means the rate limiting information
245 * for one "ip object" is shared - and these ICMPs are twice limited:
246 * by source and by destination.
247 *
248 * RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
249 * SHOULD allow setting of rate limits
250 *
251 * Shared between ICMPv4 and ICMPv6.
252 */
253#define XRLIM_BURST_FACTOR 6
254int xrlim_allow(struct dst_entry *dst, int timeout)
255{
Eric Dumazet69a73822008-01-22 06:18:34 -0800256 unsigned long now, token = dst->rate_tokens;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 int rc = 0;
258
259 now = jiffies;
Eric Dumazet69a73822008-01-22 06:18:34 -0800260 token += now - dst->rate_last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 dst->rate_last = now;
Eric Dumazet69a73822008-01-22 06:18:34 -0800262 if (token > XRLIM_BURST_FACTOR * timeout)
263 token = XRLIM_BURST_FACTOR * timeout;
264 if (token >= timeout) {
265 token -= timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 rc = 1;
267 }
Eric Dumazet69a73822008-01-22 06:18:34 -0800268 dst->rate_tokens = token;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 return rc;
270}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000271EXPORT_SYMBOL(xrlim_allow);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700273static inline int icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
274 int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275{
Changli Gaod8d1f302010-06-10 23:31:35 -0700276 struct dst_entry *dst = &rt->dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 int rc = 1;
278
279 if (type > NR_ICMP_TYPES)
280 goto out;
281
282 /* Don't limit PMTU discovery. */
283 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
284 goto out;
285
286 /* No rate limit on loopback */
287 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900288 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
290 /* Limit if icmp type is enabled in ratemask. */
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700291 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask)
292 rc = xrlim_allow(dst, net->ipv4.sysctl_icmp_ratelimit);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293out:
294 return rc;
295}
296
297/*
298 * Maintain the counters used in the SNMP statistics for outgoing ICMP
299 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700300void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700302 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700303 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304}
305
306/*
307 * Checksum each fragment, and on the first include the headers and final
308 * checksum.
309 */
310static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
311 struct sk_buff *skb)
312{
313 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800314 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315
316 csum = skb_copy_and_csum_bits(icmp_param->skb,
317 icmp_param->offset + offset,
318 to, len, 0);
319
320 skb->csum = csum_block_add(skb->csum, csum, odd);
321 if (icmp_pointers[icmp_param->data.icmph.type].error)
322 nf_ct_attach(skb, icmp_param->skb);
323 return 0;
324}
325
326static void icmp_push_reply(struct icmp_bxm *icmp_param,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800327 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800329 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 struct sk_buff *skb;
331
Changli Gaod8d1f302010-06-10 23:31:35 -0700332 sk = icmp_sk(dev_net((*rt)->dst.dev));
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800333 if (ip_append_data(sk, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900334 icmp_param->data_len+icmp_param->head_len,
335 icmp_param->head_len,
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700336 ipc, rt, MSG_DONTWAIT) < 0) {
337 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800338 ip_flush_pending_frames(sk);
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700339 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300340 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800341 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 struct sk_buff *skb1;
343
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800344 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 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;
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800352 ip_push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 }
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{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000363 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700364 struct net *net = dev_net(rt->dst.dev);
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700365 struct sock *sk;
366 struct inet_sock *inet;
Al Viro3ca3c682006-09-27 18:28:07 -0700367 __be32 daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368
369 if (ip_options_echo(&icmp_param->replyopts, skb))
Hormsf00c4012006-02-02 17:03:18 -0800370 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700372 sk = icmp_xmit_lock(net);
373 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700375 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376
377 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700379 inet->tos = ip_hdr(skb)->tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 daddr = ipc.addr = rt->rt_src;
381 ipc.opt = NULL;
Patrick Ohly51f31ca2009-02-12 05:03:39 +0000382 ipc.shtx.flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 if (icmp_param->replyopts.optlen) {
384 ipc.opt = &icmp_param->replyopts;
385 if (ipc.opt->srr)
386 daddr = icmp_param->replyopts.faddr;
387 }
388 {
389 struct flowi fl = { .nl_u = { .ip4_u =
390 { .daddr = daddr,
391 .saddr = rt->rt_spec_dst,
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700392 .tos = RT_TOS(ip_hdr(skb)->tos) } },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 .proto = IPPROTO_ICMP };
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -0700394 security_skb_classify_flow(skb, &fl);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800395 if (ip_route_output_key(net, &rt, &fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 goto out_unlock;
397 }
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700398 if (icmpv4_xrlim_allow(net, rt, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 icmp_param->data.icmph.code))
Eric Dumazet2e77d892008-11-24 15:52:46 -0800400 icmp_push_reply(icmp_param, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 ip_rt_put(rt);
402out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800403 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404}
405
406
407/*
408 * Send an ICMP message in response to a situation
409 *
410 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
411 * MAY send more (we do).
412 * MUST NOT change this header information.
413 * MUST NOT reply to a multicast/broadcast IP address.
414 * MUST NOT reply to a multicast/broadcast MAC address.
415 * MUST reply to only the first fragment.
416 */
417
Al Viroe4883012006-09-26 22:17:28 -0700418void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419{
420 struct iphdr *iph;
421 int room;
422 struct icmp_bxm icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000423 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 struct ipcm_cookie ipc;
Al Viroa61ced52006-09-26 21:27:54 -0700425 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 u8 tos;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800427 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800428 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429
430 if (!rt)
431 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700432 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
434 /*
435 * Find the original header. It is expected to be valid, of course.
436 * Check this, icmp_send is called from the most obscure devices
437 * sometimes.
438 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700439 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700441 if ((u8 *)iph < skb_in->head ||
442 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 goto out;
444
445 /*
446 * No replies to physical multicast/broadcast
447 */
448 if (skb_in->pkt_type != PACKET_HOST)
449 goto out;
450
451 /*
452 * Now check at the protocol level
453 */
454 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
455 goto out;
456
457 /*
458 * Only reply to fragment 0. We byte re-order the constant
459 * mask for efficiency.
460 */
461 if (iph->frag_off & htons(IP_OFFSET))
462 goto out;
463
464 /*
465 * If we send an ICMP error to an ICMP error a mess would result..
466 */
467 if (icmp_pointers[type].error) {
468 /*
469 * We are an error, check if we are replying to an
470 * ICMP error
471 */
472 if (iph->protocol == IPPROTO_ICMP) {
473 u8 _inner_type, *itp;
474
475 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700476 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 (iph->ihl << 2) +
478 offsetof(struct icmphdr,
479 type) -
480 skb_in->data,
481 sizeof(_inner_type),
482 &_inner_type);
483 if (itp == NULL)
484 goto out;
485
486 /*
487 * Assume any unknown ICMP type is an error. This
488 * isn't specified by the RFC, but think about it..
489 */
490 if (*itp > NR_ICMP_TYPES ||
491 icmp_pointers[*itp].error)
492 goto out;
493 }
494 }
495
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700496 sk = icmp_xmit_lock(net);
497 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 return;
499
500 /*
501 * Construct source address and options.
502 */
503
504 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700505 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700506 struct net_device *dev = NULL;
507
Eric Dumazet685c7942009-11-01 19:31:03 +0000508 rcu_read_lock();
Pavel Emelyanova24022e2008-03-26 01:55:37 -0700509 if (rt->fl.iif &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700510 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
Eric Dumazet685c7942009-11-01 19:31:03 +0000511 dev = dev_get_by_index_rcu(net, rt->fl.iif);
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700512
Eric Dumazet685c7942009-11-01 19:31:03 +0000513 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700514 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000515 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700516 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000517 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700518 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519
520 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
521 IPTOS_PREC_INTERNETCONTROL) :
522 iph->tos;
523
524 if (ip_options_echo(&icmp_param.replyopts, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800525 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
527
528 /*
529 * Prepare data for ICMP header.
530 */
531
532 icmp_param.data.icmph.type = type;
533 icmp_param.data.icmph.code = code;
534 icmp_param.data.icmph.un.gateway = info;
535 icmp_param.data.icmph.checksum = 0;
536 icmp_param.skb = skb_in;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700537 icmp_param.offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800538 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 ipc.addr = iph->saddr;
540 ipc.opt = &icmp_param.replyopts;
Patrick Ohly51f31ca2009-02-12 05:03:39 +0000541 ipc.shtx.flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542
543 {
544 struct flowi fl = {
545 .nl_u = {
546 .ip4_u = {
547 .daddr = icmp_param.replyopts.srr ?
548 icmp_param.replyopts.faddr :
549 iph->saddr,
550 .saddr = saddr,
551 .tos = RT_TOS(tos)
552 }
553 },
554 .proto = IPPROTO_ICMP,
555 .uli_u = {
556 .icmpt = {
557 .type = type,
558 .code = code
559 }
560 }
561 };
Herbert Xu8b7817f2007-12-12 10:44:43 -0800562 int err;
563 struct rtable *rt2;
564
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -0700565 security_skb_classify_flow(skb_in, &fl);
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800566 if (__ip_route_output_key(net, &rt, &fl))
Herbert Xu8b7817f2007-12-12 10:44:43 -0800567 goto out_unlock;
568
569 /* No need to clone since we're just using its address. */
570 rt2 = rt;
571
Alexey Dobriyan52479b62008-11-25 17:35:18 -0800572 err = xfrm_lookup(net, (struct dst_entry **)&rt, &fl, NULL, 0);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800573 switch (err) {
574 case 0:
575 if (rt != rt2)
576 goto route_done;
577 break;
578 case -EPERM:
579 rt = NULL;
580 break;
581 default:
582 goto out_unlock;
583 }
584
585 if (xfrm_decode_session_reverse(skb_in, &fl, AF_INET))
Herbert Xuaf268182008-04-03 12:52:19 -0700586 goto relookup_failed;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800587
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800588 if (inet_addr_type(net, fl.fl4_src) == RTN_LOCAL)
589 err = __ip_route_output_key(net, &rt2, &fl);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800590 else {
591 struct flowi fl2 = {};
Eric Dumazet7fee2262010-05-11 23:19:48 +0000592 unsigned long orefdst;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800593
594 fl2.fl4_dst = fl.fl4_src;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800595 if (ip_route_output_key(net, &rt2, &fl2))
Herbert Xuaf268182008-04-03 12:52:19 -0700596 goto relookup_failed;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800597
598 /* Ugh! */
Eric Dumazet7fee2262010-05-11 23:19:48 +0000599 orefdst = skb_in->_skb_refdst; /* save old refdst */
Herbert Xu8b7817f2007-12-12 10:44:43 -0800600 err = ip_route_input(skb_in, fl.fl4_dst, fl.fl4_src,
Changli Gaod8d1f302010-06-10 23:31:35 -0700601 RT_TOS(tos), rt2->dst.dev);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800602
Changli Gaod8d1f302010-06-10 23:31:35 -0700603 dst_release(&rt2->dst);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000604 rt2 = skb_rtable(skb_in);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000605 skb_in->_skb_refdst = orefdst; /* restore old refdst */
Herbert Xu8b7817f2007-12-12 10:44:43 -0800606 }
607
608 if (err)
Herbert Xuaf268182008-04-03 12:52:19 -0700609 goto relookup_failed;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800610
Alexey Dobriyan52479b62008-11-25 17:35:18 -0800611 err = xfrm_lookup(net, (struct dst_entry **)&rt2, &fl, NULL,
Herbert Xu8b7817f2007-12-12 10:44:43 -0800612 XFRM_LOOKUP_ICMP);
Herbert Xuaf268182008-04-03 12:52:19 -0700613 switch (err) {
614 case 0:
Changli Gaod8d1f302010-06-10 23:31:35 -0700615 dst_release(&rt->dst);
Herbert Xuaf268182008-04-03 12:52:19 -0700616 rt = rt2;
617 break;
618 case -EPERM:
619 goto ende;
620 default:
621relookup_failed:
Herbert Xu8b7817f2007-12-12 10:44:43 -0800622 if (!rt)
623 goto out_unlock;
Herbert Xuaf268182008-04-03 12:52:19 -0700624 break;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800625 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 }
627
Herbert Xu8b7817f2007-12-12 10:44:43 -0800628route_done:
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700629 if (!icmpv4_xrlim_allow(net, rt, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 goto ende;
631
632 /* RFC says return as much as we can without exceeding 576 bytes. */
633
Changli Gaod8d1f302010-06-10 23:31:35 -0700634 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 if (room > 576)
636 room = 576;
637 room -= sizeof(struct iphdr) + icmp_param.replyopts.optlen;
638 room -= sizeof(struct icmphdr);
639
640 icmp_param.data_len = skb_in->len - icmp_param.offset;
641 if (icmp_param.data_len > room)
642 icmp_param.data_len = room;
643 icmp_param.head_len = sizeof(struct icmphdr);
644
Eric Dumazet2e77d892008-11-24 15:52:46 -0800645 icmp_push_reply(&icmp_param, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646ende:
647 ip_rt_put(rt);
648out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800649 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650out:;
651}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000652EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
654
655/*
656 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
657 */
658
659static void icmp_unreach(struct sk_buff *skb)
660{
661 struct iphdr *iph;
662 struct icmphdr *icmph;
663 int hash, protocol;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000664 const struct net_protocol *ipprot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800666 struct net *net;
667
Eric Dumazetadf30902009-06-02 05:19:30 +0000668 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
670 /*
671 * Incomplete header ?
672 * Only checks for the IP header, there should be an
673 * additional check for longer headers in upper levels.
674 */
675
676 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
677 goto out_err;
678
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300679 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 iph = (struct iphdr *)skb->data;
681
682 if (iph->ihl < 5) /* Mangled header, drop. */
683 goto out_err;
684
685 if (icmph->type == ICMP_DEST_UNREACH) {
686 switch (icmph->code & 15) {
687 case ICMP_NET_UNREACH:
688 case ICMP_HOST_UNREACH:
689 case ICMP_PROT_UNREACH:
690 case ICMP_PORT_UNREACH:
691 break;
692 case ICMP_FRAG_NEEDED:
693 if (ipv4_config.no_pmtu_disc) {
Harvey Harrison673d57e2008-10-31 00:53:57 -0700694 LIMIT_NETDEBUG(KERN_INFO "ICMP: %pI4: fragmentation needed and DF set.\n",
695 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 } else {
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800697 info = ip_rt_frag_needed(net, iph,
Timo Teras0010e462008-04-29 03:32:25 -0700698 ntohs(icmph->un.frag.mtu),
699 skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 if (!info)
701 goto out;
702 }
703 break;
704 case ICMP_SR_FAILED:
Harvey Harrison673d57e2008-10-31 00:53:57 -0700705 LIMIT_NETDEBUG(KERN_INFO "ICMP: %pI4: Source Route Failed.\n",
706 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 break;
708 default:
709 break;
710 }
711 if (icmph->code > NR_ICMP_UNREACH)
712 goto out;
713 } else if (icmph->type == ICMP_PARAMETERPROB)
714 info = ntohl(icmph->un.gateway) >> 24;
715
716 /*
717 * Throw it at our lower layers
718 *
719 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
720 * header.
721 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
722 * transport layer.
723 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
724 * transport layer.
725 */
726
727 /*
728 * Check the other end isnt violating RFC 1122. Some routers send
729 * bogus responses to broadcast frames. If you see this message
730 * first check your netmask matches at both ends, if it does then
731 * get the other vendor to fix their kit.
732 */
733
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700734 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800735 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 if (net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700737 printk(KERN_WARNING "%pI4 sent an invalid ICMP "
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 "type %u, code %u "
Harvey Harrison673d57e2008-10-31 00:53:57 -0700739 "error to a broadcast: %pI4 on %s\n",
740 &ip_hdr(skb)->saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 icmph->type, icmph->code,
Harvey Harrison673d57e2008-10-31 00:53:57 -0700742 &iph->daddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 skb->dev->name);
744 goto out;
745 }
746
747 /* Checkin full IP header plus 8 bytes of protocol to
748 * avoid additional coding at protocol handlers.
749 */
750 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
751 goto out;
752
753 iph = (struct iphdr *)skb->data;
754 protocol = iph->protocol;
755
756 /*
757 * Deliver ICMP message to raw sockets. Pretty useless feature?
758 */
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800759 raw_icmp_error(skb, protocol, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 hash = protocol & (MAX_INET_PROTOS - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 rcu_read_lock();
763 ipprot = rcu_dereference(inet_protos[hash]);
764 if (ipprot && ipprot->err_handler)
765 ipprot->err_handler(skb, info);
766 rcu_read_unlock();
767
768out:
769 return;
770out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700771 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 goto out;
773}
774
775
776/*
777 * Handle ICMP_REDIRECT.
778 */
779
780static void icmp_redirect(struct sk_buff *skb)
781{
782 struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
784 if (skb->len < sizeof(struct iphdr))
785 goto out_err;
786
787 /*
788 * Get the copied header of the packet that caused the redirect
789 */
790 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
791 goto out;
792
793 iph = (struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300795 switch (icmp_hdr(skb)->code & 7) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 case ICMP_REDIR_NET:
797 case ICMP_REDIR_NETTOS:
798 /*
799 * As per RFC recommendations now handle it as a host redirect.
800 */
801 case ICMP_REDIR_HOST:
802 case ICMP_REDIR_HOSTTOS:
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700803 ip_rt_redirect(ip_hdr(skb)->saddr, iph->daddr,
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300804 icmp_hdr(skb)->un.gateway,
Ilia Sotnikovcef26852006-03-25 01:38:55 -0800805 iph->saddr, skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900807 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808out:
809 return;
810out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700811 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 goto out;
813}
814
815/*
816 * Handle ICMP_ECHO ("ping") requests.
817 *
818 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
819 * requests.
820 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
821 * included in the reply.
822 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
823 * echo requests, MUST have default=NOT.
824 * See also WRT handling of options once they are done and working.
825 */
826
827static void icmp_echo(struct sk_buff *skb)
828{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700829 struct net *net;
830
Eric Dumazetadf30902009-06-02 05:19:30 +0000831 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700832 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 struct icmp_bxm icmp_param;
834
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300835 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
837 icmp_param.skb = skb;
838 icmp_param.offset = 0;
839 icmp_param.data_len = skb->len;
840 icmp_param.head_len = sizeof(struct icmphdr);
841 icmp_reply(&icmp_param, skb);
842 }
843}
844
845/*
846 * Handle ICMP Timestamp requests.
847 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
848 * SHOULD be in the kernel for minimum random latency.
849 * MUST be accurate to a few minutes.
850 * MUST be updated at least at 15Hz.
851 */
852static void icmp_timestamp(struct sk_buff *skb)
853{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700854 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 struct icmp_bxm icmp_param;
856 /*
857 * Too short.
858 */
859 if (skb->len < 4)
860 goto out_err;
861
862 /*
863 * Fill in the current time as ms since midnight UT:
864 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700865 getnstimeofday(&tv);
866 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
867 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 icmp_param.data.times[2] = icmp_param.data.times[1];
869 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
870 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300871 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
873 icmp_param.data.icmph.code = 0;
874 icmp_param.skb = skb;
875 icmp_param.offset = 0;
876 icmp_param.data_len = 0;
877 icmp_param.head_len = sizeof(struct icmphdr) + 12;
878 icmp_reply(&icmp_param, skb);
879out:
880 return;
881out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000882 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 goto out;
884}
885
886
887/*
888 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
889 *
890 * RFC1122 (3.2.2.9). A host MUST only send replies to
891 * ADDRESS_MASK requests if it's been configured as an address mask
892 * agent. Receiving a request doesn't constitute implicit permission to
893 * act as one. Of course, implementing this correctly requires (SHOULD)
894 * a way to turn the functionality on and off. Another one for sysctl(),
895 * I guess. -- MS
896 *
897 * RFC1812 (4.3.3.9). A router MUST implement it.
898 * A router SHOULD have switch turning it on/off.
899 * This switch MUST be ON by default.
900 *
901 * Gratuitous replies, zero-source replies are not implemented,
902 * that complies with RFC. DO NOT implement them!!! All the idea
903 * of broadcast addrmask replies as specified in RFC950 is broken.
904 * The problem is that it is not uncommon to have several prefixes
905 * on one physical interface. Moreover, addrmask agent can even be
906 * not aware of existing another prefixes.
907 * If source is zero, addrmask agent cannot choose correct prefix.
908 * Gratuitous mask announcements suffer from the same problem.
909 * RFC1812 explains it, but still allows to use ADDRMASK,
910 * that is pretty silly. --ANK
911 *
912 * All these rules are so bizarre, that I removed kernel addrmask
913 * support at all. It is wrong, it is obsolete, nobody uses it in
914 * any case. --ANK
915 *
916 * Furthermore you can do it with a usermode address agent program
917 * anyway...
918 */
919
920static void icmp_address(struct sk_buff *skb)
921{
922#if 0
923 if (net_ratelimit())
924 printk(KERN_DEBUG "a guy asks for address mask. Who is it?\n");
925#endif
926}
927
928/*
929 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
930 * loudly if an inconsistency is found.
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000931 * called with rcu_read_lock()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 */
933
934static void icmp_address_reply(struct sk_buff *skb)
935{
Eric Dumazet511c3f92009-06-02 05:14:27 +0000936 struct rtable *rt = skb_rtable(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 struct net_device *dev = skb->dev;
938 struct in_device *in_dev;
939 struct in_ifaddr *ifa;
940
941 if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC))
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000942 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000944 in_dev = __in_dev_get_rcu(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 if (!in_dev)
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000946 return;
947
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 if (in_dev->ifa_list &&
949 IN_DEV_LOG_MARTIANS(in_dev) &&
950 IN_DEV_FORWARD(in_dev)) {
Al Viroa144ea42006-09-28 18:00:55 -0700951 __be32 _mask, *mp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952
953 mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask);
Kris Katterjohn09a62662006-01-08 22:24:28 -0800954 BUG_ON(mp == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
956 if (*mp == ifa->ifa_mask &&
957 inet_ifa_match(rt->rt_src, ifa))
958 break;
959 }
960 if (!ifa && net_ratelimit()) {
Harvey Harrison673d57e2008-10-31 00:53:57 -0700961 printk(KERN_INFO "Wrong address mask %pI4 from %s/%pI4\n",
962 mp, dev->name, &rt->rt_src);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 }
964 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965}
966
967static void icmp_discard(struct sk_buff *skb)
968{
969}
970
971/*
972 * Deal with incoming ICMP packets.
973 */
974int icmp_rcv(struct sk_buff *skb)
975{
976 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000977 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700978 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979
Herbert Xuaebcf822007-12-12 18:54:16 -0800980 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700981 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800982 int nh;
983
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700984 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800985 XFRM_STATE_ICMP))
986 goto drop;
987
Herbert Xu8b7817f2007-12-12 10:44:43 -0800988 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
989 goto drop;
990
991 nh = skb_network_offset(skb);
992 skb_set_network_header(skb, sizeof(*icmph));
993
994 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
995 goto drop;
996
997 skb_set_network_header(skb, nh);
998 }
999
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001000 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001
1002 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -07001003 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -08001004 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 break;
Herbert Xufb286bb2005-11-10 13:01:24 -08001006 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -08001008 skb->csum = 0;
1009 if (__skb_checksum_complete(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 }
1012
Herbert Xu8cf22942008-02-05 03:15:50 -08001013 if (!pskb_pull(skb, sizeof(*icmph)))
1014 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -03001016 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
Pavel Emelyanovf66ac032008-07-14 23:04:00 -07001018 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 /*
1020 * 18 is the highest 'known' ICMP type. Anything else is a mystery
1021 *
1022 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
1023 * discarded.
1024 */
1025 if (icmph->type > NR_ICMP_TYPES)
1026 goto error;
1027
1028
1029 /*
1030 * Parse the ICMP message
1031 */
1032
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001033 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 /*
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 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -07001040 if ((icmph->type == ICMP_ECHO ||
1041 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -07001042 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 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;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001050 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 }
1052
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 icmp_pointers[icmph->type].handler(skb);
1054
1055drop:
1056 kfree_skb(skb);
1057 return 0;
1058error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001059 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 goto drop;
1061}
1062
1063/*
1064 * This table is the definition of how we handle ICMP.
1065 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -08001066static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 [ICMP_ECHOREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 .handler = icmp_discard,
1069 },
1070 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 .handler = icmp_discard,
1072 .error = 1,
1073 },
1074 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 .handler = icmp_discard,
1076 .error = 1,
1077 },
1078 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 .handler = icmp_unreach,
1080 .error = 1,
1081 },
1082 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 .handler = icmp_unreach,
1084 .error = 1,
1085 },
1086 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 .handler = icmp_redirect,
1088 .error = 1,
1089 },
1090 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 .handler = icmp_discard,
1092 .error = 1,
1093 },
1094 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 .handler = icmp_discard,
1096 .error = 1,
1097 },
1098 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 .handler = icmp_echo,
1100 },
1101 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 .handler = icmp_discard,
1103 .error = 1,
1104 },
1105 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 .handler = icmp_discard,
1107 .error = 1,
1108 },
1109 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 .handler = icmp_unreach,
1111 .error = 1,
1112 },
1113 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 .handler = icmp_unreach,
1115 .error = 1,
1116 },
1117 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 .handler = icmp_timestamp,
1119 },
1120 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 .handler = icmp_discard,
1122 },
1123 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 .handler = icmp_discard,
1125 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001126 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 .handler = icmp_discard,
1128 },
1129 [ICMP_ADDRESS] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 .handler = icmp_address,
1131 },
1132 [ICMP_ADDRESSREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 .handler = icmp_address_reply,
1134 },
1135};
1136
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001137static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 int i;
1140
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001141 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001142 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001143 kfree(net->ipv4.icmp_sk);
1144 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001145}
1146
Adrian Bunk263173a2008-04-21 02:31:23 -07001147static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001148{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001149 int i, err;
1150
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001151 net->ipv4.icmp_sk =
1152 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1153 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001154 return -ENOMEM;
1155
Denis V. Luneva5710d62008-02-29 11:14:50 -08001156 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001157 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001159 err = inet_ctl_sock_create(&sk, PF_INET,
1160 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001162 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001164 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165
1166 /* Enough space for 2 64K ICMP packets, including
1167 * sk_buff struct overhead.
1168 */
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001169 sk->sk_sndbuf =
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 (2 * ((64 * 1024) + sizeof(struct sk_buff)));
1171
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001172 /*
1173 * Speedup sock_wfree()
1174 */
1175 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001176 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001178
1179 /* Control parameters for ECHO replies. */
1180 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1181 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1182
1183 /* Control parameter - ignore bogus broadcast responses? */
1184 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1185
1186 /*
1187 * Configurable global rate limit.
1188 *
1189 * ratelimit defines tokens/packet consumed for dst->rate_token
1190 * bucket ratemask defines which icmp types are ratelimited by
1191 * setting it's bit position.
1192 *
1193 * default:
1194 * dest unreachable (3), source quench (4),
1195 * time exceeded (11), parameter problem (12)
1196 */
1197
1198 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1199 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1200 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1201
Denis V. Luneva5710d62008-02-29 11:14:50 -08001202 return 0;
1203
1204fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001205 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001206 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001207 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001208 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209}
1210
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001211static struct pernet_operations __net_initdata icmp_sk_ops = {
1212 .init = icmp_sk_init,
1213 .exit = icmp_sk_exit,
1214};
1215
1216int __init icmp_init(void)
1217{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001218 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001219}