blob: 23ef31baa1afb84d3b81af7ab031c88830efe0f5 [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>
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +000086#include <net/ping.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070087#include <linux/skbuff.h>
88#include <net/sock.h>
89#include <linux/errno.h>
90#include <linux/timer.h>
91#include <linux/init.h>
92#include <asm/system.h>
93#include <asm/uaccess.h>
94#include <net/checksum.h>
Herbert Xu8b7817f2007-12-12 10:44:43 -080095#include <net/xfrm.h>
Denis V. Lunevc1e98942008-04-03 14:29:00 -070096#include <net/inet_common.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
98/*
99 * Build xmit assembly blocks
100 */
101
102struct icmp_bxm {
103 struct sk_buff *skb;
104 int offset;
105 int data_len;
106
107 struct {
108 struct icmphdr icmph;
Al Virob03d73e2006-09-27 18:38:30 -0700109 __be32 times[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110 } data;
111 int head_len;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000112 struct ip_options_data replyopts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113};
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
David S. Miller92d86822011-02-04 15:55:25 -0800236static inline bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
David S. Millera48eff12011-05-18 18:42:43 -0400237 struct flowi4 *fl4, int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238{
Changli Gaod8d1f302010-06-10 23:31:35 -0700239 struct dst_entry *dst = &rt->dst;
David S. Miller92d86822011-02-04 15:55:25 -0800240 bool rc = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
242 if (type > NR_ICMP_TYPES)
243 goto out;
244
245 /* Don't limit PMTU discovery. */
246 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
247 goto out;
248
249 /* No rate limit on loopback */
250 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900251 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
253 /* Limit if icmp type is enabled in ratemask. */
David S. Miller92d86822011-02-04 15:55:25 -0800254 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask) {
255 if (!rt->peer)
David S. Millera48eff12011-05-18 18:42:43 -0400256 rt_bind_peer(rt, fl4->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800257 rc = inet_peer_xrlim_allow(rt->peer,
258 net->ipv4.sysctl_icmp_ratelimit);
259 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260out:
261 return rc;
262}
263
264/*
265 * Maintain the counters used in the SNMP statistics for outgoing ICMP
266 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700267void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700269 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700270 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271}
272
273/*
274 * Checksum each fragment, and on the first include the headers and final
275 * checksum.
276 */
277static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
278 struct sk_buff *skb)
279{
280 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800281 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282
283 csum = skb_copy_and_csum_bits(icmp_param->skb,
284 icmp_param->offset + offset,
285 to, len, 0);
286
287 skb->csum = csum_block_add(skb->csum, csum, odd);
288 if (icmp_pointers[icmp_param->data.icmph.type].error)
289 nf_ct_attach(skb, icmp_param->skb);
290 return 0;
291}
292
293static void icmp_push_reply(struct icmp_bxm *icmp_param,
David S. Miller77968b72011-05-08 17:12:19 -0700294 struct flowi4 *fl4,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800295 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800297 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 struct sk_buff *skb;
299
Changli Gaod8d1f302010-06-10 23:31:35 -0700300 sk = icmp_sk(dev_net((*rt)->dst.dev));
David S. Millerf5fca602011-05-08 17:24:10 -0700301 if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900302 icmp_param->data_len+icmp_param->head_len,
303 icmp_param->head_len,
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700304 ipc, rt, MSG_DONTWAIT) < 0) {
305 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800306 ip_flush_pending_frames(sk);
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700307 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300308 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800309 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 struct sk_buff *skb1;
311
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800312 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 csum = csum_add(csum, skb1->csum);
314 }
315 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
316 (char *)icmph,
317 icmp_param->head_len, csum);
318 icmph->checksum = csum_fold(csum);
319 skb->ip_summed = CHECKSUM_NONE;
David S. Miller77968b72011-05-08 17:12:19 -0700320 ip_push_pending_frames(sk, fl4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 }
322}
323
324/*
325 * Driving logic for building and sending ICMP messages.
326 */
327
328static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
329{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000331 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700332 struct net *net = dev_net(rt->dst.dev);
David S. Miller77968b72011-05-08 17:12:19 -0700333 struct flowi4 fl4;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700334 struct sock *sk;
335 struct inet_sock *inet;
Al Viro3ca3c682006-09-27 18:28:07 -0700336 __be32 daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000338 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb))
Hormsf00c4012006-02-02 17:03:18 -0800339 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700341 sk = icmp_xmit_lock(net);
342 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700344 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345
346 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700348 inet->tos = ip_hdr(skb)->tos;
David S. Miller9f6abb52011-05-09 13:28:22 -0700349 daddr = ipc.addr = ip_hdr(skb)->saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000351 ipc.tx_flags = 0;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000352 if (icmp_param->replyopts.opt.opt.optlen) {
353 ipc.opt = &icmp_param->replyopts.opt;
354 if (ipc.opt->opt.srr)
355 daddr = icmp_param->replyopts.opt.opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 }
David S. Miller77968b72011-05-08 17:12:19 -0700357 memset(&fl4, 0, sizeof(fl4));
358 fl4.daddr = daddr;
359 fl4.saddr = rt->rt_spec_dst;
360 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
361 fl4.flowi4_proto = IPPROTO_ICMP;
362 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
363 rt = ip_route_output_key(net, &fl4);
364 if (IS_ERR(rt))
365 goto out_unlock;
David S. Millera48eff12011-05-18 18:42:43 -0400366 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 icmp_param->data.icmph.code))
David S. Miller77968b72011-05-08 17:12:19 -0700368 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 ip_rt_put(rt);
370out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800371 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
373
David S. Miller77968b72011-05-08 17:12:19 -0700374static struct rtable *icmp_route_lookup(struct net *net,
375 struct flowi4 *fl4,
376 struct sk_buff *skb_in,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000377 const struct iphdr *iph,
David S. Millerf6d460c2011-03-01 15:49:55 -0800378 __be32 saddr, u8 tos,
379 int type, int code,
380 struct icmp_bxm *param)
381{
David S. Millerf6d460c2011-03-01 15:49:55 -0800382 struct rtable *rt, *rt2;
David S. Miller0671b302011-07-22 06:22:10 -0700383 struct flowi4 fl4_dec;
David S. Millerf6d460c2011-03-01 15:49:55 -0800384 int err;
385
David S. Miller77968b72011-05-08 17:12:19 -0700386 memset(fl4, 0, sizeof(*fl4));
387 fl4->daddr = (param->replyopts.opt.opt.srr ?
388 param->replyopts.opt.opt.faddr : iph->saddr);
389 fl4->saddr = saddr;
390 fl4->flowi4_tos = RT_TOS(tos);
391 fl4->flowi4_proto = IPPROTO_ICMP;
392 fl4->fl4_icmp_type = type;
393 fl4->fl4_icmp_code = code;
394 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
395 rt = __ip_route_output_key(net, fl4);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800396 if (IS_ERR(rt))
397 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800398
399 /* No need to clone since we're just using its address. */
400 rt2 = rt;
401
David S. Miller9d6ec932011-03-12 01:12:47 -0500402 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
David S. Miller77968b72011-05-08 17:12:19 -0700403 flowi4_to_flowi(fl4), NULL, 0);
David S. Miller452edd52011-03-02 13:27:41 -0800404 if (!IS_ERR(rt)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800405 if (rt != rt2)
406 return rt;
David S. Miller452edd52011-03-02 13:27:41 -0800407 } else if (PTR_ERR(rt) == -EPERM) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800408 rt = NULL;
David S. Miller452edd52011-03-02 13:27:41 -0800409 } else
410 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800411
David S. Miller0671b302011-07-22 06:22:10 -0700412 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
David S. Millerf6d460c2011-03-01 15:49:55 -0800413 if (err)
414 goto relookup_failed;
415
David S. Miller0671b302011-07-22 06:22:10 -0700416 if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) {
417 rt2 = __ip_route_output_key(net, &fl4_dec);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800418 if (IS_ERR(rt2))
419 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800420 } else {
David S. Miller9d6ec932011-03-12 01:12:47 -0500421 struct flowi4 fl4_2 = {};
David S. Millerf6d460c2011-03-01 15:49:55 -0800422 unsigned long orefdst;
423
David S. Miller0671b302011-07-22 06:22:10 -0700424 fl4_2.daddr = fl4_dec.saddr;
David S. Miller9d6ec932011-03-12 01:12:47 -0500425 rt2 = ip_route_output_key(net, &fl4_2);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800426 if (IS_ERR(rt2)) {
427 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800428 goto relookup_failed;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800429 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800430 /* Ugh! */
431 orefdst = skb_in->_skb_refdst; /* save old refdst */
David S. Miller0671b302011-07-22 06:22:10 -0700432 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
David S. Millerf6d460c2011-03-01 15:49:55 -0800433 RT_TOS(tos), rt2->dst.dev);
434
435 dst_release(&rt2->dst);
436 rt2 = skb_rtable(skb_in);
437 skb_in->_skb_refdst = orefdst; /* restore old refdst */
438 }
439
440 if (err)
441 goto relookup_failed;
442
David S. Miller9d6ec932011-03-12 01:12:47 -0500443 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
David S. Miller0671b302011-07-22 06:22:10 -0700444 flowi4_to_flowi(&fl4_dec), NULL,
David S. Miller9d6ec932011-03-12 01:12:47 -0500445 XFRM_LOOKUP_ICMP);
David S. Miller452edd52011-03-02 13:27:41 -0800446 if (!IS_ERR(rt2)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800447 dst_release(&rt->dst);
David S. Miller0671b302011-07-22 06:22:10 -0700448 memcpy(fl4, &fl4_dec, sizeof(*fl4));
David S. Millerf6d460c2011-03-01 15:49:55 -0800449 rt = rt2;
David S. Miller452edd52011-03-02 13:27:41 -0800450 } else if (PTR_ERR(rt2) == -EPERM) {
451 if (rt)
452 dst_release(&rt->dst);
453 return rt2;
454 } else {
455 err = PTR_ERR(rt2);
456 goto relookup_failed;
David S. Millerf6d460c2011-03-01 15:49:55 -0800457 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800458 return rt;
459
460relookup_failed:
461 if (rt)
462 return rt;
463 return ERR_PTR(err);
464}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466/*
467 * Send an ICMP message in response to a situation
468 *
469 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
470 * MAY send more (we do).
471 * MUST NOT change this header information.
472 * MUST NOT reply to a multicast/broadcast IP address.
473 * MUST NOT reply to a multicast/broadcast MAC address.
474 * MUST reply to only the first fragment.
475 */
476
Al Viroe4883012006-09-26 22:17:28 -0700477void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478{
479 struct iphdr *iph;
480 int room;
481 struct icmp_bxm icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000482 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 struct ipcm_cookie ipc;
David S. Miller77968b72011-05-08 17:12:19 -0700484 struct flowi4 fl4;
Al Viroa61ced52006-09-26 21:27:54 -0700485 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 u8 tos;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800487 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800488 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489
490 if (!rt)
491 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700492 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
494 /*
495 * Find the original header. It is expected to be valid, of course.
496 * Check this, icmp_send is called from the most obscure devices
497 * sometimes.
498 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700499 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700501 if ((u8 *)iph < skb_in->head ||
502 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 goto out;
504
505 /*
506 * No replies to physical multicast/broadcast
507 */
508 if (skb_in->pkt_type != PACKET_HOST)
509 goto out;
510
511 /*
512 * Now check at the protocol level
513 */
514 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
515 goto out;
516
517 /*
518 * Only reply to fragment 0. We byte re-order the constant
519 * mask for efficiency.
520 */
521 if (iph->frag_off & htons(IP_OFFSET))
522 goto out;
523
524 /*
525 * If we send an ICMP error to an ICMP error a mess would result..
526 */
527 if (icmp_pointers[type].error) {
528 /*
529 * We are an error, check if we are replying to an
530 * ICMP error
531 */
532 if (iph->protocol == IPPROTO_ICMP) {
533 u8 _inner_type, *itp;
534
535 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700536 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 (iph->ihl << 2) +
538 offsetof(struct icmphdr,
539 type) -
540 skb_in->data,
541 sizeof(_inner_type),
542 &_inner_type);
543 if (itp == NULL)
544 goto out;
545
546 /*
547 * Assume any unknown ICMP type is an error. This
548 * isn't specified by the RFC, but think about it..
549 */
550 if (*itp > NR_ICMP_TYPES ||
551 icmp_pointers[*itp].error)
552 goto out;
553 }
554 }
555
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700556 sk = icmp_xmit_lock(net);
557 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 return;
559
560 /*
561 * Construct source address and options.
562 */
563
564 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700565 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700566 struct net_device *dev = NULL;
567
Eric Dumazet685c7942009-11-01 19:31:03 +0000568 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800569 if (rt_is_input_route(rt) &&
570 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller5e2b61f2011-03-04 21:47:09 -0800571 dev = dev_get_by_index_rcu(net, rt->rt_iif);
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700572
Eric Dumazet685c7942009-11-01 19:31:03 +0000573 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700574 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000575 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700576 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000577 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700578 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579
580 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
581 IPTOS_PREC_INTERNETCONTROL) :
582 iph->tos;
583
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000584 if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800585 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586
587
588 /*
589 * Prepare data for ICMP header.
590 */
591
592 icmp_param.data.icmph.type = type;
593 icmp_param.data.icmph.code = code;
594 icmp_param.data.icmph.un.gateway = info;
595 icmp_param.data.icmph.checksum = 0;
596 icmp_param.skb = skb_in;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700597 icmp_param.offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800598 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 ipc.addr = iph->saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000600 ipc.opt = &icmp_param.replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000601 ipc.tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
David S. Miller77968b72011-05-08 17:12:19 -0700603 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos,
David S. Millerf6d460c2011-03-01 15:49:55 -0800604 type, code, &icmp_param);
605 if (IS_ERR(rt))
606 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800607
David S. Millera48eff12011-05-18 18:42:43 -0400608 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 goto ende;
610
611 /* RFC says return as much as we can without exceeding 576 bytes. */
612
Changli Gaod8d1f302010-06-10 23:31:35 -0700613 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 if (room > 576)
615 room = 576;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000616 room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 room -= sizeof(struct icmphdr);
618
619 icmp_param.data_len = skb_in->len - icmp_param.offset;
620 if (icmp_param.data_len > room)
621 icmp_param.data_len = room;
622 icmp_param.head_len = sizeof(struct icmphdr);
623
David S. Miller77968b72011-05-08 17:12:19 -0700624 icmp_push_reply(&icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625ende:
626 ip_rt_put(rt);
627out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800628 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629out:;
630}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000631EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632
633
634/*
635 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
636 */
637
638static void icmp_unreach(struct sk_buff *skb)
639{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000640 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 struct icmphdr *icmph;
642 int hash, protocol;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000643 const struct net_protocol *ipprot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800645 struct net *net;
646
Eric Dumazetadf30902009-06-02 05:19:30 +0000647 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
649 /*
650 * Incomplete header ?
651 * Only checks for the IP header, there should be an
652 * additional check for longer headers in upper levels.
653 */
654
655 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
656 goto out_err;
657
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300658 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000659 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
661 if (iph->ihl < 5) /* Mangled header, drop. */
662 goto out_err;
663
664 if (icmph->type == ICMP_DEST_UNREACH) {
665 switch (icmph->code & 15) {
666 case ICMP_NET_UNREACH:
667 case ICMP_HOST_UNREACH:
668 case ICMP_PROT_UNREACH:
669 case ICMP_PORT_UNREACH:
670 break;
671 case ICMP_FRAG_NEEDED:
672 if (ipv4_config.no_pmtu_disc) {
Harvey Harrison673d57e2008-10-31 00:53:57 -0700673 LIMIT_NETDEBUG(KERN_INFO "ICMP: %pI4: fragmentation needed and DF set.\n",
674 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 } else {
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800676 info = ip_rt_frag_needed(net, iph,
Timo Teras0010e462008-04-29 03:32:25 -0700677 ntohs(icmph->un.frag.mtu),
678 skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 if (!info)
680 goto out;
681 }
682 break;
683 case ICMP_SR_FAILED:
Harvey Harrison673d57e2008-10-31 00:53:57 -0700684 LIMIT_NETDEBUG(KERN_INFO "ICMP: %pI4: Source Route Failed.\n",
685 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 break;
687 default:
688 break;
689 }
690 if (icmph->code > NR_ICMP_UNREACH)
691 goto out;
692 } else if (icmph->type == ICMP_PARAMETERPROB)
693 info = ntohl(icmph->un.gateway) >> 24;
694
695 /*
696 * Throw it at our lower layers
697 *
698 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
699 * header.
700 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
701 * transport layer.
702 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
703 * transport layer.
704 */
705
706 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300707 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 * bogus responses to broadcast frames. If you see this message
709 * first check your netmask matches at both ends, if it does then
710 * get the other vendor to fix their kit.
711 */
712
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700713 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800714 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 if (net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700716 printk(KERN_WARNING "%pI4 sent an invalid ICMP "
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 "type %u, code %u "
Harvey Harrison673d57e2008-10-31 00:53:57 -0700718 "error to a broadcast: %pI4 on %s\n",
719 &ip_hdr(skb)->saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 icmph->type, icmph->code,
Harvey Harrison673d57e2008-10-31 00:53:57 -0700721 &iph->daddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 skb->dev->name);
723 goto out;
724 }
725
726 /* Checkin full IP header plus 8 bytes of protocol to
727 * avoid additional coding at protocol handlers.
728 */
729 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
730 goto out;
731
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000732 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 protocol = iph->protocol;
734
735 /*
736 * Deliver ICMP message to raw sockets. Pretty useless feature?
737 */
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800738 raw_icmp_error(skb, protocol, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 hash = protocol & (MAX_INET_PROTOS - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 rcu_read_lock();
742 ipprot = rcu_dereference(inet_protos[hash]);
743 if (ipprot && ipprot->err_handler)
744 ipprot->err_handler(skb, info);
745 rcu_read_unlock();
746
747out:
748 return;
749out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700750 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 goto out;
752}
753
754
755/*
756 * Handle ICMP_REDIRECT.
757 */
758
759static void icmp_redirect(struct sk_buff *skb)
760{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000761 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
763 if (skb->len < sizeof(struct iphdr))
764 goto out_err;
765
766 /*
767 * Get the copied header of the packet that caused the redirect
768 */
769 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
770 goto out;
771
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000772 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300774 switch (icmp_hdr(skb)->code & 7) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 case ICMP_REDIR_NET:
776 case ICMP_REDIR_NETTOS:
777 /*
778 * As per RFC recommendations now handle it as a host redirect.
779 */
780 case ICMP_REDIR_HOST:
781 case ICMP_REDIR_HOSTTOS:
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700782 ip_rt_redirect(ip_hdr(skb)->saddr, iph->daddr,
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300783 icmp_hdr(skb)->un.gateway,
Ilia Sotnikovcef26852006-03-25 01:38:55 -0800784 iph->saddr, skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900786 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000787
788 /* Ping wants to see redirects.
789 * Let's pretend they are errors of sorts... */
790 if (iph->protocol == IPPROTO_ICMP &&
791 iph->ihl >= 5 &&
792 pskb_may_pull(skb, (iph->ihl<<2)+8)) {
793 ping_err(skb, icmp_hdr(skb)->un.gateway);
794 }
795
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796out:
797 return;
798out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700799 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 goto out;
801}
802
803/*
804 * Handle ICMP_ECHO ("ping") requests.
805 *
806 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
807 * requests.
808 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
809 * included in the reply.
810 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
811 * echo requests, MUST have default=NOT.
812 * See also WRT handling of options once they are done and working.
813 */
814
815static void icmp_echo(struct sk_buff *skb)
816{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700817 struct net *net;
818
Eric Dumazetadf30902009-06-02 05:19:30 +0000819 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700820 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 struct icmp_bxm icmp_param;
822
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300823 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
825 icmp_param.skb = skb;
826 icmp_param.offset = 0;
827 icmp_param.data_len = skb->len;
828 icmp_param.head_len = sizeof(struct icmphdr);
829 icmp_reply(&icmp_param, skb);
830 }
831}
832
833/*
834 * Handle ICMP Timestamp requests.
835 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
836 * SHOULD be in the kernel for minimum random latency.
837 * MUST be accurate to a few minutes.
838 * MUST be updated at least at 15Hz.
839 */
840static void icmp_timestamp(struct sk_buff *skb)
841{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700842 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 struct icmp_bxm icmp_param;
844 /*
845 * Too short.
846 */
847 if (skb->len < 4)
848 goto out_err;
849
850 /*
851 * Fill in the current time as ms since midnight UT:
852 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700853 getnstimeofday(&tv);
854 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
855 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 icmp_param.data.times[2] = icmp_param.data.times[1];
857 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
858 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300859 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
861 icmp_param.data.icmph.code = 0;
862 icmp_param.skb = skb;
863 icmp_param.offset = 0;
864 icmp_param.data_len = 0;
865 icmp_param.head_len = sizeof(struct icmphdr) + 12;
866 icmp_reply(&icmp_param, skb);
867out:
868 return;
869out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000870 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 goto out;
872}
873
874
875/*
876 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
877 *
878 * RFC1122 (3.2.2.9). A host MUST only send replies to
879 * ADDRESS_MASK requests if it's been configured as an address mask
880 * agent. Receiving a request doesn't constitute implicit permission to
881 * act as one. Of course, implementing this correctly requires (SHOULD)
882 * a way to turn the functionality on and off. Another one for sysctl(),
883 * I guess. -- MS
884 *
885 * RFC1812 (4.3.3.9). A router MUST implement it.
886 * A router SHOULD have switch turning it on/off.
887 * This switch MUST be ON by default.
888 *
889 * Gratuitous replies, zero-source replies are not implemented,
890 * that complies with RFC. DO NOT implement them!!! All the idea
891 * of broadcast addrmask replies as specified in RFC950 is broken.
892 * The problem is that it is not uncommon to have several prefixes
893 * on one physical interface. Moreover, addrmask agent can even be
894 * not aware of existing another prefixes.
895 * If source is zero, addrmask agent cannot choose correct prefix.
896 * Gratuitous mask announcements suffer from the same problem.
897 * RFC1812 explains it, but still allows to use ADDRMASK,
898 * that is pretty silly. --ANK
899 *
900 * All these rules are so bizarre, that I removed kernel addrmask
901 * support at all. It is wrong, it is obsolete, nobody uses it in
902 * any case. --ANK
903 *
904 * Furthermore you can do it with a usermode address agent program
905 * anyway...
906 */
907
908static void icmp_address(struct sk_buff *skb)
909{
910#if 0
911 if (net_ratelimit())
912 printk(KERN_DEBUG "a guy asks for address mask. Who is it?\n");
913#endif
914}
915
916/*
917 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
918 * loudly if an inconsistency is found.
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000919 * called with rcu_read_lock()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 */
921
922static void icmp_address_reply(struct sk_buff *skb)
923{
Eric Dumazet511c3f92009-06-02 05:14:27 +0000924 struct rtable *rt = skb_rtable(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 struct net_device *dev = skb->dev;
926 struct in_device *in_dev;
927 struct in_ifaddr *ifa;
928
929 if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC))
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000930 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000932 in_dev = __in_dev_get_rcu(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 if (!in_dev)
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000934 return;
935
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 if (in_dev->ifa_list &&
937 IN_DEV_LOG_MARTIANS(in_dev) &&
938 IN_DEV_FORWARD(in_dev)) {
Al Viroa144ea42006-09-28 18:00:55 -0700939 __be32 _mask, *mp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940
941 mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask);
Kris Katterjohn09a62662006-01-08 22:24:28 -0800942 BUG_ON(mp == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
944 if (*mp == ifa->ifa_mask &&
David S. Miller9f6abb52011-05-09 13:28:22 -0700945 inet_ifa_match(ip_hdr(skb)->saddr, ifa))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 break;
947 }
948 if (!ifa && net_ratelimit()) {
Harvey Harrison673d57e2008-10-31 00:53:57 -0700949 printk(KERN_INFO "Wrong address mask %pI4 from %s/%pI4\n",
David S. Miller9f6abb52011-05-09 13:28:22 -0700950 mp, dev->name, &ip_hdr(skb)->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 }
952 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953}
954
955static void icmp_discard(struct sk_buff *skb)
956{
957}
958
959/*
960 * Deal with incoming ICMP packets.
961 */
962int icmp_rcv(struct sk_buff *skb)
963{
964 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000965 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700966 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967
Herbert Xuaebcf822007-12-12 18:54:16 -0800968 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700969 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800970 int nh;
971
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700972 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800973 XFRM_STATE_ICMP))
974 goto drop;
975
Herbert Xu8b7817f2007-12-12 10:44:43 -0800976 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
977 goto drop;
978
979 nh = skb_network_offset(skb);
980 skb_set_network_header(skb, sizeof(*icmph));
981
982 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
983 goto drop;
984
985 skb_set_network_header(skb, nh);
986 }
987
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700988 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989
990 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700991 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800992 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 break;
Herbert Xufb286bb2005-11-10 13:01:24 -0800994 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -0800996 skb->csum = 0;
997 if (__skb_checksum_complete(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 }
1000
Herbert Xu8cf22942008-02-05 03:15:50 -08001001 if (!pskb_pull(skb, sizeof(*icmph)))
1002 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -03001004 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
Pavel Emelyanovf66ac032008-07-14 23:04:00 -07001006 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 /*
1008 * 18 is the highest 'known' ICMP type. Anything else is a mystery
1009 *
1010 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
1011 * discarded.
1012 */
1013 if (icmph->type > NR_ICMP_TYPES)
1014 goto error;
1015
1016
1017 /*
1018 * Parse the ICMP message
1019 */
1020
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001021 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 /*
1023 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
1024 * silently ignored (we let user decide with a sysctl).
1025 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1026 * discarded if to broadcast/multicast.
1027 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -07001028 if ((icmph->type == ICMP_ECHO ||
1029 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -07001030 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 goto error;
1032 }
1033 if (icmph->type != ICMP_ECHO &&
1034 icmph->type != ICMP_TIMESTAMP &&
1035 icmph->type != ICMP_ADDRESS &&
1036 icmph->type != ICMP_ADDRESSREPLY) {
1037 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001038 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 }
1040
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 icmp_pointers[icmph->type].handler(skb);
1042
1043drop:
1044 kfree_skb(skb);
1045 return 0;
1046error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001047 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 goto drop;
1049}
1050
1051/*
1052 * This table is the definition of how we handle ICMP.
1053 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -08001054static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001056 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 },
1058 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 .handler = icmp_discard,
1060 .error = 1,
1061 },
1062 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 .handler = icmp_discard,
1064 .error = 1,
1065 },
1066 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 .handler = icmp_unreach,
1068 .error = 1,
1069 },
1070 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 .handler = icmp_unreach,
1072 .error = 1,
1073 },
1074 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 .handler = icmp_redirect,
1076 .error = 1,
1077 },
1078 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 .handler = icmp_discard,
1080 .error = 1,
1081 },
1082 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 .handler = icmp_discard,
1084 .error = 1,
1085 },
1086 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 .handler = icmp_echo,
1088 },
1089 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 .handler = icmp_discard,
1091 .error = 1,
1092 },
1093 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 .handler = icmp_discard,
1095 .error = 1,
1096 },
1097 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 .handler = icmp_unreach,
1099 .error = 1,
1100 },
1101 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 .handler = icmp_unreach,
1103 .error = 1,
1104 },
1105 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 .handler = icmp_timestamp,
1107 },
1108 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 .handler = icmp_discard,
1110 },
1111 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 .handler = icmp_discard,
1113 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001114 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 .handler = icmp_discard,
1116 },
1117 [ICMP_ADDRESS] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 .handler = icmp_address,
1119 },
1120 [ICMP_ADDRESSREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 .handler = icmp_address_reply,
1122 },
1123};
1124
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001125static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 int i;
1128
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001129 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001130 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001131 kfree(net->ipv4.icmp_sk);
1132 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001133}
1134
Adrian Bunk263173a2008-04-21 02:31:23 -07001135static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001136{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001137 int i, err;
1138
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001139 net->ipv4.icmp_sk =
1140 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1141 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001142 return -ENOMEM;
1143
Denis V. Luneva5710d62008-02-29 11:14:50 -08001144 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001145 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001147 err = inet_ctl_sock_create(&sk, PF_INET,
1148 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001150 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001152 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
1154 /* Enough space for 2 64K ICMP packets, including
1155 * sk_buff struct overhead.
1156 */
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001157 sk->sk_sndbuf =
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 (2 * ((64 * 1024) + sizeof(struct sk_buff)));
1159
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001160 /*
1161 * Speedup sock_wfree()
1162 */
1163 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001164 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001166
1167 /* Control parameters for ECHO replies. */
1168 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1169 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1170
1171 /* Control parameter - ignore bogus broadcast responses? */
1172 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1173
1174 /*
1175 * Configurable global rate limit.
1176 *
1177 * ratelimit defines tokens/packet consumed for dst->rate_token
1178 * bucket ratemask defines which icmp types are ratelimited by
1179 * setting it's bit position.
1180 *
1181 * default:
1182 * dest unreachable (3), source quench (4),
1183 * time exceeded (11), parameter problem (12)
1184 */
1185
1186 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1187 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1188 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1189
Denis V. Luneva5710d62008-02-29 11:14:50 -08001190 return 0;
1191
1192fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001193 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001194 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001195 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001196 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197}
1198
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001199static struct pernet_operations __net_initdata icmp_sk_ops = {
1200 .init = icmp_sk_init,
1201 .exit = icmp_sk_exit,
1202};
1203
1204int __init icmp_init(void)
1205{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001206 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001207}