blob: 70a793559bb54d4a1a44af13f8a9e7a7b2658708 [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
Joe Perchesafd465032012-03-12 07:03:32 +000065#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/module.h>
68#include <linux/types.h>
69#include <linux/jiffies.h>
70#include <linux/kernel.h>
71#include <linux/fcntl.h>
72#include <linux/socket.h>
73#include <linux/in.h>
74#include <linux/inet.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020075#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/netdevice.h>
77#include <linux/string.h>
78#include <linux/netfilter_ipv4.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090079#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#include <net/snmp.h>
81#include <net/ip.h>
82#include <net/route.h>
83#include <net/protocol.h>
84#include <net/icmp.h>
85#include <net/tcp.h>
86#include <net/udp.h>
87#include <net/raw.h>
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +000088#include <net/ping.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070089#include <linux/skbuff.h>
90#include <net/sock.h>
91#include <linux/errno.h>
92#include <linux/timer.h>
93#include <linux/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include <asm/uaccess.h>
95#include <net/checksum.h>
Herbert Xu8b7817f2007-12-12 10:44:43 -080096#include <net/xfrm.h>
Denis V. Lunevc1e98942008-04-03 14:29:00 -070097#include <net/inet_common.h>
David S. Miller35ebf652012-06-28 03:59:11 -070098#include <net/ip_fib.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100/*
101 * Build xmit assembly blocks
102 */
103
104struct icmp_bxm {
105 struct sk_buff *skb;
106 int offset;
107 int data_len;
108
109 struct {
110 struct icmphdr icmph;
Al Virob03d73e2006-09-27 18:38:30 -0700111 __be32 times[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 } data;
113 int head_len;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000114 struct ip_options_data replyopts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
116
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117/* An array of errno for error messages from dest unreach. */
118/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
119
Alexey Dobriyane7548342010-01-22 10:18:25 +0000120const struct icmp_err icmp_err_convert[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121 {
122 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
123 .fatal = 0,
124 },
125 {
126 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
127 .fatal = 0,
128 },
129 {
130 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
131 .fatal = 1,
132 },
133 {
134 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
135 .fatal = 1,
136 },
137 {
138 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
139 .fatal = 0,
140 },
141 {
142 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
143 .fatal = 0,
144 },
145 {
146 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
147 .fatal = 1,
148 },
149 {
150 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
151 .fatal = 1,
152 },
153 {
154 .errno = ENONET, /* ICMP_HOST_ISOLATED */
155 .fatal = 1,
156 },
157 {
158 .errno = ENETUNREACH, /* ICMP_NET_ANO */
159 .fatal = 1,
160 },
161 {
162 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
163 .fatal = 1,
164 },
165 {
166 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
167 .fatal = 0,
168 },
169 {
170 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
171 .fatal = 0,
172 },
173 {
174 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
175 .fatal = 1,
176 },
177 {
178 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
179 .fatal = 1,
180 },
181 {
182 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
183 .fatal = 1,
184 },
185};
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000186EXPORT_SYMBOL(icmp_err_convert);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188/*
189 * ICMP control array. This specifies what to do with each ICMP.
190 */
191
192struct icmp_control {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 void (*handler)(struct sk_buff *skb);
194 short error; /* This ICMP is classed as an error message */
195};
196
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800197static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198
199/*
200 * The ICMP socket(s). This is the most convenient way to flow control
201 * our ICMP output as well as maintain a clean interface throughout
202 * all layers. All Socketless IP sends will soon be gone.
203 *
204 * On SMP we have one ICMP socket per-cpu.
205 */
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800206static struct sock *icmp_sk(struct net *net)
207{
208 return net->ipv4.icmp_sk[smp_processor_id()];
209}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700211static inline struct sock *icmp_xmit_lock(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212{
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700213 struct sock *sk;
214
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 local_bh_disable();
216
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700217 sk = icmp_sk(net);
218
Denis V. Lunev405666d2008-02-29 11:16:46 -0800219 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 /* This can happen if the output path signals a
221 * dst_link_failure() for an outgoing ICMP packet.
222 */
223 local_bh_enable();
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700224 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 }
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700226 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227}
228
Denis V. Lunev405666d2008-02-29 11:16:46 -0800229static inline void icmp_xmit_unlock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230{
Denis V. Lunev405666d2008-02-29 11:16:46 -0800231 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232}
233
234/*
235 * Send an ICMP frame.
236 */
237
David S. Miller92d86822011-02-04 15:55:25 -0800238static inline bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
David S. Millera48eff12011-05-18 18:42:43 -0400239 struct flowi4 *fl4, int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240{
Changli Gaod8d1f302010-06-10 23:31:35 -0700241 struct dst_entry *dst = &rt->dst;
David S. Miller92d86822011-02-04 15:55:25 -0800242 bool rc = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
244 if (type > NR_ICMP_TYPES)
245 goto out;
246
247 /* Don't limit PMTU discovery. */
248 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
249 goto out;
250
251 /* No rate limit on loopback */
252 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900253 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
255 /* Limit if icmp type is enabled in ratemask. */
David S. Miller92d86822011-02-04 15:55:25 -0800256 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask) {
David S. Miller1d861aa2012-07-10 03:58:16 -0700257 struct inet_peer *peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, 1);
David S. Millerfbfe95a2012-06-08 23:24:18 -0700258 rc = inet_peer_xrlim_allow(peer,
David S. Miller92d86822011-02-04 15:55:25 -0800259 net->ipv4.sysctl_icmp_ratelimit);
David S. Miller1d861aa2012-07-10 03:58:16 -0700260 inet_putpeer(peer);
David S. Miller92d86822011-02-04 15:55:25 -0800261 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262out:
263 return rc;
264}
265
266/*
267 * Maintain the counters used in the SNMP statistics for outgoing ICMP
268 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700269void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700271 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700272 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273}
274
275/*
276 * Checksum each fragment, and on the first include the headers and final
277 * checksum.
278 */
279static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
280 struct sk_buff *skb)
281{
282 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800283 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
285 csum = skb_copy_and_csum_bits(icmp_param->skb,
286 icmp_param->offset + offset,
287 to, len, 0);
288
289 skb->csum = csum_block_add(skb->csum, csum, odd);
290 if (icmp_pointers[icmp_param->data.icmph.type].error)
291 nf_ct_attach(skb, icmp_param->skb);
292 return 0;
293}
294
295static void icmp_push_reply(struct icmp_bxm *icmp_param,
David S. Miller77968b72011-05-08 17:12:19 -0700296 struct flowi4 *fl4,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800297 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800299 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 struct sk_buff *skb;
301
Changli Gaod8d1f302010-06-10 23:31:35 -0700302 sk = icmp_sk(dev_net((*rt)->dst.dev));
David S. Millerf5fca602011-05-08 17:24:10 -0700303 if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900304 icmp_param->data_len+icmp_param->head_len,
305 icmp_param->head_len,
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700306 ipc, rt, MSG_DONTWAIT) < 0) {
307 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800308 ip_flush_pending_frames(sk);
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700309 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300310 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800311 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 struct sk_buff *skb1;
313
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800314 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 csum = csum_add(csum, skb1->csum);
316 }
317 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
318 (char *)icmph,
319 icmp_param->head_len, csum);
320 icmph->checksum = csum_fold(csum);
321 skb->ip_summed = CHECKSUM_NONE;
David S. Miller77968b72011-05-08 17:12:19 -0700322 ip_push_pending_frames(sk, fl4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 }
324}
325
326/*
327 * Driving logic for building and sending ICMP messages.
328 */
329
330static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
331{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000333 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700334 struct net *net = dev_net(rt->dst.dev);
David S. Miller77968b72011-05-08 17:12:19 -0700335 struct flowi4 fl4;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700336 struct sock *sk;
337 struct inet_sock *inet;
David S. Miller35ebf652012-06-28 03:59:11 -0700338 __be32 daddr, saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000340 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb))
Hormsf00c4012006-02-02 17:03:18 -0800341 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700343 sk = icmp_xmit_lock(net);
344 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700346 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
348 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700350 inet->tos = ip_hdr(skb)->tos;
David S. Miller9f6abb52011-05-09 13:28:22 -0700351 daddr = ipc.addr = ip_hdr(skb)->saddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700352 saddr = fib_compute_spec_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000354 ipc.tx_flags = 0;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000355 if (icmp_param->replyopts.opt.opt.optlen) {
356 ipc.opt = &icmp_param->replyopts.opt;
357 if (ipc.opt->opt.srr)
358 daddr = icmp_param->replyopts.opt.opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 }
David S. Miller77968b72011-05-08 17:12:19 -0700360 memset(&fl4, 0, sizeof(fl4));
361 fl4.daddr = daddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700362 fl4.saddr = saddr;
David S. Miller77968b72011-05-08 17:12:19 -0700363 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
364 fl4.flowi4_proto = IPPROTO_ICMP;
365 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
366 rt = ip_route_output_key(net, &fl4);
367 if (IS_ERR(rt))
368 goto out_unlock;
David S. Millera48eff12011-05-18 18:42:43 -0400369 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 icmp_param->data.icmph.code))
David S. Miller77968b72011-05-08 17:12:19 -0700371 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 ip_rt_put(rt);
373out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800374 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375}
376
David S. Miller77968b72011-05-08 17:12:19 -0700377static struct rtable *icmp_route_lookup(struct net *net,
378 struct flowi4 *fl4,
379 struct sk_buff *skb_in,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000380 const struct iphdr *iph,
David S. Millerf6d460c2011-03-01 15:49:55 -0800381 __be32 saddr, u8 tos,
382 int type, int code,
383 struct icmp_bxm *param)
384{
David S. Millerf6d460c2011-03-01 15:49:55 -0800385 struct rtable *rt, *rt2;
David S. Miller415b3332011-07-22 06:22:10 -0700386 struct flowi4 fl4_dec;
David S. Millerf6d460c2011-03-01 15:49:55 -0800387 int err;
388
David S. Miller77968b72011-05-08 17:12:19 -0700389 memset(fl4, 0, sizeof(*fl4));
390 fl4->daddr = (param->replyopts.opt.opt.srr ?
391 param->replyopts.opt.opt.faddr : iph->saddr);
392 fl4->saddr = saddr;
393 fl4->flowi4_tos = RT_TOS(tos);
394 fl4->flowi4_proto = IPPROTO_ICMP;
395 fl4->fl4_icmp_type = type;
396 fl4->fl4_icmp_code = code;
397 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
398 rt = __ip_route_output_key(net, fl4);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800399 if (IS_ERR(rt))
400 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800401
402 /* No need to clone since we're just using its address. */
403 rt2 = rt;
404
David S. Miller9d6ec932011-03-12 01:12:47 -0500405 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
David S. Miller77968b72011-05-08 17:12:19 -0700406 flowi4_to_flowi(fl4), NULL, 0);
David S. Miller452edd52011-03-02 13:27:41 -0800407 if (!IS_ERR(rt)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800408 if (rt != rt2)
409 return rt;
David S. Miller452edd52011-03-02 13:27:41 -0800410 } else if (PTR_ERR(rt) == -EPERM) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800411 rt = NULL;
David S. Miller452edd52011-03-02 13:27:41 -0800412 } else
413 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800414
David S. Miller415b3332011-07-22 06:22:10 -0700415 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
David S. Millerf6d460c2011-03-01 15:49:55 -0800416 if (err)
417 goto relookup_failed;
418
David S. Miller415b3332011-07-22 06:22:10 -0700419 if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) {
420 rt2 = __ip_route_output_key(net, &fl4_dec);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800421 if (IS_ERR(rt2))
422 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800423 } else {
David S. Miller9d6ec932011-03-12 01:12:47 -0500424 struct flowi4 fl4_2 = {};
David S. Millerf6d460c2011-03-01 15:49:55 -0800425 unsigned long orefdst;
426
David S. Miller415b3332011-07-22 06:22:10 -0700427 fl4_2.daddr = fl4_dec.saddr;
David S. Miller9d6ec932011-03-12 01:12:47 -0500428 rt2 = ip_route_output_key(net, &fl4_2);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800429 if (IS_ERR(rt2)) {
430 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800431 goto relookup_failed;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800432 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800433 /* Ugh! */
434 orefdst = skb_in->_skb_refdst; /* save old refdst */
David S. Miller415b3332011-07-22 06:22:10 -0700435 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
David S. Millerf6d460c2011-03-01 15:49:55 -0800436 RT_TOS(tos), rt2->dst.dev);
437
438 dst_release(&rt2->dst);
439 rt2 = skb_rtable(skb_in);
440 skb_in->_skb_refdst = orefdst; /* restore old refdst */
441 }
442
443 if (err)
444 goto relookup_failed;
445
David S. Miller9d6ec932011-03-12 01:12:47 -0500446 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
David S. Miller415b3332011-07-22 06:22:10 -0700447 flowi4_to_flowi(&fl4_dec), NULL,
David S. Miller9d6ec932011-03-12 01:12:47 -0500448 XFRM_LOOKUP_ICMP);
David S. Miller452edd52011-03-02 13:27:41 -0800449 if (!IS_ERR(rt2)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800450 dst_release(&rt->dst);
David S. Miller415b3332011-07-22 06:22:10 -0700451 memcpy(fl4, &fl4_dec, sizeof(*fl4));
David S. Millerf6d460c2011-03-01 15:49:55 -0800452 rt = rt2;
David S. Miller452edd52011-03-02 13:27:41 -0800453 } else if (PTR_ERR(rt2) == -EPERM) {
454 if (rt)
455 dst_release(&rt->dst);
456 return rt2;
457 } else {
458 err = PTR_ERR(rt2);
459 goto relookup_failed;
David S. Millerf6d460c2011-03-01 15:49:55 -0800460 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800461 return rt;
462
463relookup_failed:
464 if (rt)
465 return rt;
466 return ERR_PTR(err);
467}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
469/*
470 * Send an ICMP message in response to a situation
471 *
472 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
473 * MAY send more (we do).
474 * MUST NOT change this header information.
475 * MUST NOT reply to a multicast/broadcast IP address.
476 * MUST NOT reply to a multicast/broadcast MAC address.
477 * MUST reply to only the first fragment.
478 */
479
Al Viroe4883012006-09-26 22:17:28 -0700480void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481{
482 struct iphdr *iph;
483 int room;
484 struct icmp_bxm icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000485 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 struct ipcm_cookie ipc;
David S. Miller77968b72011-05-08 17:12:19 -0700487 struct flowi4 fl4;
Al Viroa61ced52006-09-26 21:27:54 -0700488 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 u8 tos;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800490 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800491 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492
493 if (!rt)
494 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700495 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
497 /*
498 * Find the original header. It is expected to be valid, of course.
499 * Check this, icmp_send is called from the most obscure devices
500 * sometimes.
501 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700502 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700504 if ((u8 *)iph < skb_in->head ||
505 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 goto out;
507
508 /*
509 * No replies to physical multicast/broadcast
510 */
511 if (skb_in->pkt_type != PACKET_HOST)
512 goto out;
513
514 /*
515 * Now check at the protocol level
516 */
517 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
518 goto out;
519
520 /*
521 * Only reply to fragment 0. We byte re-order the constant
522 * mask for efficiency.
523 */
524 if (iph->frag_off & htons(IP_OFFSET))
525 goto out;
526
527 /*
528 * If we send an ICMP error to an ICMP error a mess would result..
529 */
530 if (icmp_pointers[type].error) {
531 /*
532 * We are an error, check if we are replying to an
533 * ICMP error
534 */
535 if (iph->protocol == IPPROTO_ICMP) {
536 u8 _inner_type, *itp;
537
538 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700539 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 (iph->ihl << 2) +
541 offsetof(struct icmphdr,
542 type) -
543 skb_in->data,
544 sizeof(_inner_type),
545 &_inner_type);
546 if (itp == NULL)
547 goto out;
548
549 /*
550 * Assume any unknown ICMP type is an error. This
551 * isn't specified by the RFC, but think about it..
552 */
553 if (*itp > NR_ICMP_TYPES ||
554 icmp_pointers[*itp].error)
555 goto out;
556 }
557 }
558
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700559 sk = icmp_xmit_lock(net);
560 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 return;
562
563 /*
564 * Construct source address and options.
565 */
566
567 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700568 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700569 struct net_device *dev = NULL;
570
Eric Dumazet685c7942009-11-01 19:31:03 +0000571 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800572 if (rt_is_input_route(rt) &&
573 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller5e2b61f2011-03-04 21:47:09 -0800574 dev = dev_get_by_index_rcu(net, rt->rt_iif);
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700575
Eric Dumazet685c7942009-11-01 19:31:03 +0000576 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700577 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000578 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700579 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000580 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700581 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
583 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
584 IPTOS_PREC_INTERNETCONTROL) :
585 iph->tos;
586
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000587 if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800588 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589
590
591 /*
592 * Prepare data for ICMP header.
593 */
594
595 icmp_param.data.icmph.type = type;
596 icmp_param.data.icmph.code = code;
597 icmp_param.data.icmph.un.gateway = info;
598 icmp_param.data.icmph.checksum = 0;
599 icmp_param.skb = skb_in;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700600 icmp_param.offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800601 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 ipc.addr = iph->saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000603 ipc.opt = &icmp_param.replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000604 ipc.tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
David S. Miller77968b72011-05-08 17:12:19 -0700606 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos,
David S. Millerf6d460c2011-03-01 15:49:55 -0800607 type, code, &icmp_param);
608 if (IS_ERR(rt))
609 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800610
David S. Millera48eff12011-05-18 18:42:43 -0400611 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 goto ende;
613
614 /* RFC says return as much as we can without exceeding 576 bytes. */
615
Changli Gaod8d1f302010-06-10 23:31:35 -0700616 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 if (room > 576)
618 room = 576;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000619 room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 room -= sizeof(struct icmphdr);
621
622 icmp_param.data_len = skb_in->len - icmp_param.offset;
623 if (icmp_param.data_len > room)
624 icmp_param.data_len = room;
625 icmp_param.head_len = sizeof(struct icmphdr);
626
David S. Miller77968b72011-05-08 17:12:19 -0700627 icmp_push_reply(&icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628ende:
629 ip_rt_put(rt);
630out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800631 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632out:;
633}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000634EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635
636
David S. Miller1de92432012-07-11 18:32:17 -0700637static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
638{
639 const struct iphdr *iph = (const struct iphdr *) skb->data;
640 const struct net_protocol *ipprot;
641 int protocol = iph->protocol;
642
643 raw_icmp_error(skb, protocol, info);
644
645 rcu_read_lock();
646 ipprot = rcu_dereference(inet_protos[protocol]);
647 if (ipprot && ipprot->err_handler)
648 ipprot->err_handler(skb, info);
649 rcu_read_unlock();
650}
651
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652/*
653 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
654 */
655
656static void icmp_unreach(struct sk_buff *skb)
657{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000658 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 struct icmphdr *icmph;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800660 struct net *net;
David S. Millerf9242b62012-06-19 18:56:21 -0700661 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800662
Eric Dumazetadf30902009-06-02 05:19:30 +0000663 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
665 /*
666 * Incomplete header ?
667 * Only checks for the IP header, there should be an
668 * additional check for longer headers in upper levels.
669 */
670
671 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
672 goto out_err;
673
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300674 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000675 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
677 if (iph->ihl < 5) /* Mangled header, drop. */
678 goto out_err;
679
680 if (icmph->type == ICMP_DEST_UNREACH) {
681 switch (icmph->code & 15) {
682 case ICMP_NET_UNREACH:
683 case ICMP_HOST_UNREACH:
684 case ICMP_PROT_UNREACH:
685 case ICMP_PORT_UNREACH:
686 break;
687 case ICMP_FRAG_NEEDED:
688 if (ipv4_config.no_pmtu_disc) {
Joe Perchesafd465032012-03-12 07:03:32 +0000689 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700690 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 } else {
David S. Miller46517002012-06-10 00:04:12 -0700692 info = ntohs(icmph->un.frag.mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 if (!info)
694 goto out;
695 }
696 break;
697 case ICMP_SR_FAILED:
Joe Perchesafd465032012-03-12 07:03:32 +0000698 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700699 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 break;
701 default:
702 break;
703 }
704 if (icmph->code > NR_ICMP_UNREACH)
705 goto out;
706 } else if (icmph->type == ICMP_PARAMETERPROB)
707 info = ntohl(icmph->un.gateway) >> 24;
708
709 /*
710 * Throw it at our lower layers
711 *
712 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
713 * header.
714 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
715 * transport layer.
716 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
717 * transport layer.
718 */
719
720 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300721 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 * bogus responses to broadcast frames. If you see this message
723 * first check your netmask matches at both ends, if it does then
724 * get the other vendor to fix their kit.
725 */
726
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700727 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800728 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Joe Perchese87cc472012-05-13 21:56:26 +0000729 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
730 &ip_hdr(skb)->saddr,
731 icmph->type, icmph->code,
732 &iph->daddr, skb->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 goto out;
734 }
735
736 /* Checkin full IP header plus 8 bytes of protocol to
737 * avoid additional coding at protocol handlers.
738 */
739 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
740 goto out;
741
David S. Miller1de92432012-07-11 18:32:17 -0700742 icmp_socket_deliver(skb, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
744out:
745 return;
746out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700747 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 goto out;
749}
750
751
752/*
753 * Handle ICMP_REDIRECT.
754 */
755
756static void icmp_redirect(struct sk_buff *skb)
757{
David S. Miller94206122012-07-11 20:38:08 -0700758 if (skb->len < sizeof(struct iphdr)) {
759 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
760 return;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900761 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000762
David S. Miller94206122012-07-11 20:38:08 -0700763 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
764 return;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000765
David S. Miller94206122012-07-11 20:38:08 -0700766 ip_rt_redirect(skb, icmp_hdr(skb)->un.gateway);
767 icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768}
769
770/*
771 * Handle ICMP_ECHO ("ping") requests.
772 *
773 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
774 * requests.
775 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
776 * included in the reply.
777 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
778 * echo requests, MUST have default=NOT.
779 * See also WRT handling of options once they are done and working.
780 */
781
782static void icmp_echo(struct sk_buff *skb)
783{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700784 struct net *net;
785
Eric Dumazetadf30902009-06-02 05:19:30 +0000786 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700787 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 struct icmp_bxm icmp_param;
789
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300790 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
792 icmp_param.skb = skb;
793 icmp_param.offset = 0;
794 icmp_param.data_len = skb->len;
795 icmp_param.head_len = sizeof(struct icmphdr);
796 icmp_reply(&icmp_param, skb);
797 }
798}
799
800/*
801 * Handle ICMP Timestamp requests.
802 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
803 * SHOULD be in the kernel for minimum random latency.
804 * MUST be accurate to a few minutes.
805 * MUST be updated at least at 15Hz.
806 */
807static void icmp_timestamp(struct sk_buff *skb)
808{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700809 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 struct icmp_bxm icmp_param;
811 /*
812 * Too short.
813 */
814 if (skb->len < 4)
815 goto out_err;
816
817 /*
818 * Fill in the current time as ms since midnight UT:
819 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700820 getnstimeofday(&tv);
821 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
822 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 icmp_param.data.times[2] = icmp_param.data.times[1];
824 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
825 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300826 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
828 icmp_param.data.icmph.code = 0;
829 icmp_param.skb = skb;
830 icmp_param.offset = 0;
831 icmp_param.data_len = 0;
832 icmp_param.head_len = sizeof(struct icmphdr) + 12;
833 icmp_reply(&icmp_param, skb);
834out:
835 return;
836out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000837 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 goto out;
839}
840
841
842/*
843 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
844 *
845 * RFC1122 (3.2.2.9). A host MUST only send replies to
846 * ADDRESS_MASK requests if it's been configured as an address mask
847 * agent. Receiving a request doesn't constitute implicit permission to
848 * act as one. Of course, implementing this correctly requires (SHOULD)
849 * a way to turn the functionality on and off. Another one for sysctl(),
850 * I guess. -- MS
851 *
852 * RFC1812 (4.3.3.9). A router MUST implement it.
853 * A router SHOULD have switch turning it on/off.
854 * This switch MUST be ON by default.
855 *
856 * Gratuitous replies, zero-source replies are not implemented,
857 * that complies with RFC. DO NOT implement them!!! All the idea
858 * of broadcast addrmask replies as specified in RFC950 is broken.
859 * The problem is that it is not uncommon to have several prefixes
860 * on one physical interface. Moreover, addrmask agent can even be
861 * not aware of existing another prefixes.
862 * If source is zero, addrmask agent cannot choose correct prefix.
863 * Gratuitous mask announcements suffer from the same problem.
864 * RFC1812 explains it, but still allows to use ADDRMASK,
865 * that is pretty silly. --ANK
866 *
867 * All these rules are so bizarre, that I removed kernel addrmask
868 * support at all. It is wrong, it is obsolete, nobody uses it in
869 * any case. --ANK
870 *
871 * Furthermore you can do it with a usermode address agent program
872 * anyway...
873 */
874
875static void icmp_address(struct sk_buff *skb)
876{
877#if 0
Joe Perchese87cc472012-05-13 21:56:26 +0000878 net_dbg_ratelimited("a guy asks for address mask. Who is it?\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879#endif
880}
881
882/*
883 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
884 * loudly if an inconsistency is found.
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000885 * called with rcu_read_lock()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 */
887
888static void icmp_address_reply(struct sk_buff *skb)
889{
Eric Dumazet511c3f92009-06-02 05:14:27 +0000890 struct rtable *rt = skb_rtable(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 struct net_device *dev = skb->dev;
892 struct in_device *in_dev;
893 struct in_ifaddr *ifa;
894
895 if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC))
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000896 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000898 in_dev = __in_dev_get_rcu(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 if (!in_dev)
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000900 return;
901
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 if (in_dev->ifa_list &&
903 IN_DEV_LOG_MARTIANS(in_dev) &&
904 IN_DEV_FORWARD(in_dev)) {
Al Viroa144ea42006-09-28 18:00:55 -0700905 __be32 _mask, *mp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
907 mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask);
Kris Katterjohn09a62662006-01-08 22:24:28 -0800908 BUG_ON(mp == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
910 if (*mp == ifa->ifa_mask &&
David S. Miller9f6abb52011-05-09 13:28:22 -0700911 inet_ifa_match(ip_hdr(skb)->saddr, ifa))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 break;
913 }
Joe Perchese87cc472012-05-13 21:56:26 +0000914 if (!ifa)
915 net_info_ratelimited("Wrong address mask %pI4 from %s/%pI4\n",
916 mp,
917 dev->name, &ip_hdr(skb)->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919}
920
921static void icmp_discard(struct sk_buff *skb)
922{
923}
924
925/*
926 * Deal with incoming ICMP packets.
927 */
928int icmp_rcv(struct sk_buff *skb)
929{
930 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000931 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700932 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933
Herbert Xuaebcf822007-12-12 18:54:16 -0800934 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700935 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800936 int nh;
937
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700938 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800939 XFRM_STATE_ICMP))
940 goto drop;
941
Herbert Xu8b7817f2007-12-12 10:44:43 -0800942 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
943 goto drop;
944
945 nh = skb_network_offset(skb);
946 skb_set_network_header(skb, sizeof(*icmph));
947
948 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
949 goto drop;
950
951 skb_set_network_header(skb, nh);
952 }
953
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700954 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
956 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700957 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800958 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 break;
Herbert Xufb286bb2005-11-10 13:01:24 -0800960 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -0800962 skb->csum = 0;
963 if (__skb_checksum_complete(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 }
966
Herbert Xu8cf22942008-02-05 03:15:50 -0800967 if (!pskb_pull(skb, sizeof(*icmph)))
968 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300970 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971
Pavel Emelyanovf66ac032008-07-14 23:04:00 -0700972 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 /*
974 * 18 is the highest 'known' ICMP type. Anything else is a mystery
975 *
976 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
977 * discarded.
978 */
979 if (icmph->type > NR_ICMP_TYPES)
980 goto error;
981
982
983 /*
984 * Parse the ICMP message
985 */
986
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900987 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 /*
989 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
990 * silently ignored (we let user decide with a sysctl).
991 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
992 * discarded if to broadcast/multicast.
993 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -0700994 if ((icmph->type == ICMP_ECHO ||
995 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700996 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 goto error;
998 }
999 if (icmph->type != ICMP_ECHO &&
1000 icmph->type != ICMP_TIMESTAMP &&
1001 icmph->type != ICMP_ADDRESS &&
1002 icmph->type != ICMP_ADDRESSREPLY) {
1003 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001004 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 }
1006
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 icmp_pointers[icmph->type].handler(skb);
1008
1009drop:
1010 kfree_skb(skb);
1011 return 0;
1012error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001013 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 goto drop;
1015}
1016
1017/*
1018 * This table is the definition of how we handle ICMP.
1019 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -08001020static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001022 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 },
1024 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 .handler = icmp_discard,
1026 .error = 1,
1027 },
1028 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 .handler = icmp_discard,
1030 .error = 1,
1031 },
1032 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 .handler = icmp_unreach,
1034 .error = 1,
1035 },
1036 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 .handler = icmp_unreach,
1038 .error = 1,
1039 },
1040 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 .handler = icmp_redirect,
1042 .error = 1,
1043 },
1044 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 .handler = icmp_discard,
1046 .error = 1,
1047 },
1048 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 .handler = icmp_discard,
1050 .error = 1,
1051 },
1052 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 .handler = icmp_echo,
1054 },
1055 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 .handler = icmp_discard,
1057 .error = 1,
1058 },
1059 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 .handler = icmp_discard,
1061 .error = 1,
1062 },
1063 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 .handler = icmp_unreach,
1065 .error = 1,
1066 },
1067 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 .handler = icmp_unreach,
1069 .error = 1,
1070 },
1071 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 .handler = icmp_timestamp,
1073 },
1074 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 .handler = icmp_discard,
1076 },
1077 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 .handler = icmp_discard,
1079 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001080 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 .handler = icmp_discard,
1082 },
1083 [ICMP_ADDRESS] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 .handler = icmp_address,
1085 },
1086 [ICMP_ADDRESSREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 .handler = icmp_address_reply,
1088 },
1089};
1090
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001091static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 int i;
1094
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001095 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001096 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001097 kfree(net->ipv4.icmp_sk);
1098 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001099}
1100
Adrian Bunk263173a2008-04-21 02:31:23 -07001101static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001102{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001103 int i, err;
1104
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001105 net->ipv4.icmp_sk =
1106 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1107 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001108 return -ENOMEM;
1109
Denis V. Luneva5710d62008-02-29 11:14:50 -08001110 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001111 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001113 err = inet_ctl_sock_create(&sk, PF_INET,
1114 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001116 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001118 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119
1120 /* Enough space for 2 64K ICMP packets, including
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001121 * sk_buff/skb_shared_info struct overhead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 */
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001123 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001125 /*
1126 * Speedup sock_wfree()
1127 */
1128 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001129 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001131
1132 /* Control parameters for ECHO replies. */
1133 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1134 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1135
1136 /* Control parameter - ignore bogus broadcast responses? */
1137 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1138
1139 /*
1140 * Configurable global rate limit.
1141 *
1142 * ratelimit defines tokens/packet consumed for dst->rate_token
1143 * bucket ratemask defines which icmp types are ratelimited by
1144 * setting it's bit position.
1145 *
1146 * default:
1147 * dest unreachable (3), source quench (4),
1148 * time exceeded (11), parameter problem (12)
1149 */
1150
1151 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1152 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1153 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1154
Denis V. Luneva5710d62008-02-29 11:14:50 -08001155 return 0;
1156
1157fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001158 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001159 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001160 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001161 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162}
1163
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001164static struct pernet_operations __net_initdata icmp_sk_ops = {
1165 .init = icmp_sk_init,
1166 .exit = icmp_sk_exit,
1167};
1168
1169int __init icmp_init(void)
1170{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001171 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001172}