blob: 2cb2bf845641da6bee915edfb73240c9e675c565 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99/*
100 * Build xmit assembly blocks
101 */
102
103struct icmp_bxm {
104 struct sk_buff *skb;
105 int offset;
106 int data_len;
107
108 struct {
109 struct icmphdr icmph;
Al Virob03d73e2006-09-27 18:38:30 -0700110 __be32 times[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 } data;
112 int head_len;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000113 struct ip_options_data replyopts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114};
115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116/* An array of errno for error messages from dest unreach. */
117/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
118
Alexey Dobriyane7548342010-01-22 10:18:25 +0000119const struct icmp_err icmp_err_convert[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120 {
121 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
122 .fatal = 0,
123 },
124 {
125 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
126 .fatal = 0,
127 },
128 {
129 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
130 .fatal = 1,
131 },
132 {
133 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
134 .fatal = 1,
135 },
136 {
137 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
138 .fatal = 0,
139 },
140 {
141 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
142 .fatal = 0,
143 },
144 {
145 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
146 .fatal = 1,
147 },
148 {
149 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
150 .fatal = 1,
151 },
152 {
153 .errno = ENONET, /* ICMP_HOST_ISOLATED */
154 .fatal = 1,
155 },
156 {
157 .errno = ENETUNREACH, /* ICMP_NET_ANO */
158 .fatal = 1,
159 },
160 {
161 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
162 .fatal = 1,
163 },
164 {
165 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
166 .fatal = 0,
167 },
168 {
169 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
170 .fatal = 0,
171 },
172 {
173 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
174 .fatal = 1,
175 },
176 {
177 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
178 .fatal = 1,
179 },
180 {
181 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
182 .fatal = 1,
183 },
184};
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000185EXPORT_SYMBOL(icmp_err_convert);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187/*
188 * ICMP control array. This specifies what to do with each ICMP.
189 */
190
191struct icmp_control {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 void (*handler)(struct sk_buff *skb);
193 short error; /* This ICMP is classed as an error message */
194};
195
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800196static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197
198/*
199 * The ICMP socket(s). This is the most convenient way to flow control
200 * our ICMP output as well as maintain a clean interface throughout
201 * all layers. All Socketless IP sends will soon be gone.
202 *
203 * On SMP we have one ICMP socket per-cpu.
204 */
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800205static struct sock *icmp_sk(struct net *net)
206{
207 return net->ipv4.icmp_sk[smp_processor_id()];
208}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700210static inline struct sock *icmp_xmit_lock(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211{
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700212 struct sock *sk;
213
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 local_bh_disable();
215
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700216 sk = icmp_sk(net);
217
Denis V. Lunev405666d2008-02-29 11:16:46 -0800218 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 /* This can happen if the output path signals a
220 * dst_link_failure() for an outgoing ICMP packet.
221 */
222 local_bh_enable();
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700223 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 }
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700225 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226}
227
Denis V. Lunev405666d2008-02-29 11:16:46 -0800228static inline void icmp_xmit_unlock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229{
Denis V. Lunev405666d2008-02-29 11:16:46 -0800230 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231}
232
233/*
234 * Send an ICMP frame.
235 */
236
David S. Miller92d86822011-02-04 15:55:25 -0800237static inline bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
David S. Millera48eff12011-05-18 18:42:43 -0400238 struct flowi4 *fl4, int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239{
Changli Gaod8d1f302010-06-10 23:31:35 -0700240 struct dst_entry *dst = &rt->dst;
David S. Miller92d86822011-02-04 15:55:25 -0800241 bool rc = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
243 if (type > NR_ICMP_TYPES)
244 goto out;
245
246 /* Don't limit PMTU discovery. */
247 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
248 goto out;
249
250 /* No rate limit on loopback */
251 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900252 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
254 /* Limit if icmp type is enabled in ratemask. */
David S. Miller92d86822011-02-04 15:55:25 -0800255 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask) {
256 if (!rt->peer)
David S. Millera48eff12011-05-18 18:42:43 -0400257 rt_bind_peer(rt, fl4->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800258 rc = inet_peer_xrlim_allow(rt->peer,
259 net->ipv4.sysctl_icmp_ratelimit);
260 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261out:
262 return rc;
263}
264
265/*
266 * Maintain the counters used in the SNMP statistics for outgoing ICMP
267 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700268void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700270 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700271 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272}
273
274/*
275 * Checksum each fragment, and on the first include the headers and final
276 * checksum.
277 */
278static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
279 struct sk_buff *skb)
280{
281 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800282 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283
284 csum = skb_copy_and_csum_bits(icmp_param->skb,
285 icmp_param->offset + offset,
286 to, len, 0);
287
288 skb->csum = csum_block_add(skb->csum, csum, odd);
289 if (icmp_pointers[icmp_param->data.icmph.type].error)
290 nf_ct_attach(skb, icmp_param->skb);
291 return 0;
292}
293
294static void icmp_push_reply(struct icmp_bxm *icmp_param,
David S. Miller77968b72011-05-08 17:12:19 -0700295 struct flowi4 *fl4,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800296 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800298 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 struct sk_buff *skb;
300
Changli Gaod8d1f302010-06-10 23:31:35 -0700301 sk = icmp_sk(dev_net((*rt)->dst.dev));
David S. Millerf5fca602011-05-08 17:24:10 -0700302 if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900303 icmp_param->data_len+icmp_param->head_len,
304 icmp_param->head_len,
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700305 ipc, rt, MSG_DONTWAIT) < 0) {
306 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800307 ip_flush_pending_frames(sk);
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700308 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300309 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800310 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 struct sk_buff *skb1;
312
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800313 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 csum = csum_add(csum, skb1->csum);
315 }
316 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
317 (char *)icmph,
318 icmp_param->head_len, csum);
319 icmph->checksum = csum_fold(csum);
320 skb->ip_summed = CHECKSUM_NONE;
David S. Miller77968b72011-05-08 17:12:19 -0700321 ip_push_pending_frames(sk, fl4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 }
323}
324
325/*
326 * Driving logic for building and sending ICMP messages.
327 */
328
329static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
330{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000332 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700333 struct net *net = dev_net(rt->dst.dev);
David S. Miller77968b72011-05-08 17:12:19 -0700334 struct flowi4 fl4;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700335 struct sock *sk;
336 struct inet_sock *inet;
Al Viro3ca3c682006-09-27 18:28:07 -0700337 __be32 daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000339 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb))
Hormsf00c4012006-02-02 17:03:18 -0800340 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700342 sk = icmp_xmit_lock(net);
343 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700345 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346
347 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700349 inet->tos = ip_hdr(skb)->tos;
David S. Miller9f6abb52011-05-09 13:28:22 -0700350 daddr = ipc.addr = ip_hdr(skb)->saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000352 ipc.tx_flags = 0;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000353 if (icmp_param->replyopts.opt.opt.optlen) {
354 ipc.opt = &icmp_param->replyopts.opt;
355 if (ipc.opt->opt.srr)
356 daddr = icmp_param->replyopts.opt.opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 }
David S. Miller77968b72011-05-08 17:12:19 -0700358 memset(&fl4, 0, sizeof(fl4));
359 fl4.daddr = daddr;
360 fl4.saddr = rt->rt_spec_dst;
361 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
362 fl4.flowi4_proto = IPPROTO_ICMP;
363 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
364 rt = ip_route_output_key(net, &fl4);
365 if (IS_ERR(rt))
366 goto out_unlock;
David S. Millera48eff12011-05-18 18:42:43 -0400367 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 icmp_param->data.icmph.code))
David S. Miller77968b72011-05-08 17:12:19 -0700369 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 ip_rt_put(rt);
371out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800372 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373}
374
David S. Miller77968b72011-05-08 17:12:19 -0700375static struct rtable *icmp_route_lookup(struct net *net,
376 struct flowi4 *fl4,
377 struct sk_buff *skb_in,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000378 const struct iphdr *iph,
David S. Millerf6d460c2011-03-01 15:49:55 -0800379 __be32 saddr, u8 tos,
380 int type, int code,
381 struct icmp_bxm *param)
382{
David S. Millerf6d460c2011-03-01 15:49:55 -0800383 struct rtable *rt, *rt2;
David S. Miller415b3332011-07-22 06:22:10 -0700384 struct flowi4 fl4_dec;
David S. Millerf6d460c2011-03-01 15:49:55 -0800385 int err;
386
David S. Miller77968b72011-05-08 17:12:19 -0700387 memset(fl4, 0, sizeof(*fl4));
388 fl4->daddr = (param->replyopts.opt.opt.srr ?
389 param->replyopts.opt.opt.faddr : iph->saddr);
390 fl4->saddr = saddr;
391 fl4->flowi4_tos = RT_TOS(tos);
392 fl4->flowi4_proto = IPPROTO_ICMP;
393 fl4->fl4_icmp_type = type;
394 fl4->fl4_icmp_code = code;
395 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
396 rt = __ip_route_output_key(net, fl4);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800397 if (IS_ERR(rt))
398 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800399
400 /* No need to clone since we're just using its address. */
401 rt2 = rt;
402
David S. Miller9d6ec932011-03-12 01:12:47 -0500403 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
David S. Miller77968b72011-05-08 17:12:19 -0700404 flowi4_to_flowi(fl4), NULL, 0);
David S. Miller452edd52011-03-02 13:27:41 -0800405 if (!IS_ERR(rt)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800406 if (rt != rt2)
407 return rt;
David S. Miller452edd52011-03-02 13:27:41 -0800408 } else if (PTR_ERR(rt) == -EPERM) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800409 rt = NULL;
David S. Miller452edd52011-03-02 13:27:41 -0800410 } else
411 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800412
David S. Miller415b3332011-07-22 06:22:10 -0700413 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
David S. Millerf6d460c2011-03-01 15:49:55 -0800414 if (err)
415 goto relookup_failed;
416
David S. Miller415b3332011-07-22 06:22:10 -0700417 if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) {
418 rt2 = __ip_route_output_key(net, &fl4_dec);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800419 if (IS_ERR(rt2))
420 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800421 } else {
David S. Miller9d6ec932011-03-12 01:12:47 -0500422 struct flowi4 fl4_2 = {};
David S. Millerf6d460c2011-03-01 15:49:55 -0800423 unsigned long orefdst;
424
David S. Miller415b3332011-07-22 06:22:10 -0700425 fl4_2.daddr = fl4_dec.saddr;
David S. Miller9d6ec932011-03-12 01:12:47 -0500426 rt2 = ip_route_output_key(net, &fl4_2);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800427 if (IS_ERR(rt2)) {
428 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800429 goto relookup_failed;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800430 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800431 /* Ugh! */
432 orefdst = skb_in->_skb_refdst; /* save old refdst */
David S. Miller415b3332011-07-22 06:22:10 -0700433 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
David S. Millerf6d460c2011-03-01 15:49:55 -0800434 RT_TOS(tos), rt2->dst.dev);
435
436 dst_release(&rt2->dst);
437 rt2 = skb_rtable(skb_in);
438 skb_in->_skb_refdst = orefdst; /* restore old refdst */
439 }
440
441 if (err)
442 goto relookup_failed;
443
David S. Miller9d6ec932011-03-12 01:12:47 -0500444 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
David S. Miller415b3332011-07-22 06:22:10 -0700445 flowi4_to_flowi(&fl4_dec), NULL,
David S. Miller9d6ec932011-03-12 01:12:47 -0500446 XFRM_LOOKUP_ICMP);
David S. Miller452edd52011-03-02 13:27:41 -0800447 if (!IS_ERR(rt2)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800448 dst_release(&rt->dst);
David S. Miller415b3332011-07-22 06:22:10 -0700449 memcpy(fl4, &fl4_dec, sizeof(*fl4));
David S. Millerf6d460c2011-03-01 15:49:55 -0800450 rt = rt2;
David S. Miller452edd52011-03-02 13:27:41 -0800451 } else if (PTR_ERR(rt2) == -EPERM) {
452 if (rt)
453 dst_release(&rt->dst);
454 return rt2;
455 } else {
456 err = PTR_ERR(rt2);
457 goto relookup_failed;
David S. Millerf6d460c2011-03-01 15:49:55 -0800458 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800459 return rt;
460
461relookup_failed:
462 if (rt)
463 return rt;
464 return ERR_PTR(err);
465}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
467/*
468 * Send an ICMP message in response to a situation
469 *
470 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
471 * MAY send more (we do).
472 * MUST NOT change this header information.
473 * MUST NOT reply to a multicast/broadcast IP address.
474 * MUST NOT reply to a multicast/broadcast MAC address.
475 * MUST reply to only the first fragment.
476 */
477
Al Viroe4883012006-09-26 22:17:28 -0700478void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479{
480 struct iphdr *iph;
481 int room;
482 struct icmp_bxm icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000483 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 struct ipcm_cookie ipc;
David S. Miller77968b72011-05-08 17:12:19 -0700485 struct flowi4 fl4;
Al Viroa61ced52006-09-26 21:27:54 -0700486 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 u8 tos;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800488 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800489 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
491 if (!rt)
492 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700493 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494
495 /*
496 * Find the original header. It is expected to be valid, of course.
497 * Check this, icmp_send is called from the most obscure devices
498 * sometimes.
499 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700500 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700502 if ((u8 *)iph < skb_in->head ||
503 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 goto out;
505
506 /*
507 * No replies to physical multicast/broadcast
508 */
509 if (skb_in->pkt_type != PACKET_HOST)
510 goto out;
511
512 /*
513 * Now check at the protocol level
514 */
515 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
516 goto out;
517
518 /*
519 * Only reply to fragment 0. We byte re-order the constant
520 * mask for efficiency.
521 */
522 if (iph->frag_off & htons(IP_OFFSET))
523 goto out;
524
525 /*
526 * If we send an ICMP error to an ICMP error a mess would result..
527 */
528 if (icmp_pointers[type].error) {
529 /*
530 * We are an error, check if we are replying to an
531 * ICMP error
532 */
533 if (iph->protocol == IPPROTO_ICMP) {
534 u8 _inner_type, *itp;
535
536 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700537 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 (iph->ihl << 2) +
539 offsetof(struct icmphdr,
540 type) -
541 skb_in->data,
542 sizeof(_inner_type),
543 &_inner_type);
544 if (itp == NULL)
545 goto out;
546
547 /*
548 * Assume any unknown ICMP type is an error. This
549 * isn't specified by the RFC, but think about it..
550 */
551 if (*itp > NR_ICMP_TYPES ||
552 icmp_pointers[*itp].error)
553 goto out;
554 }
555 }
556
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700557 sk = icmp_xmit_lock(net);
558 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 return;
560
561 /*
562 * Construct source address and options.
563 */
564
565 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700566 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700567 struct net_device *dev = NULL;
568
Eric Dumazet685c7942009-11-01 19:31:03 +0000569 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800570 if (rt_is_input_route(rt) &&
571 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller5e2b61f2011-03-04 21:47:09 -0800572 dev = dev_get_by_index_rcu(net, rt->rt_iif);
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700573
Eric Dumazet685c7942009-11-01 19:31:03 +0000574 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700575 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000576 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700577 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000578 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700579 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
581 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
582 IPTOS_PREC_INTERNETCONTROL) :
583 iph->tos;
584
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000585 if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800586 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587
588
589 /*
590 * Prepare data for ICMP header.
591 */
592
593 icmp_param.data.icmph.type = type;
594 icmp_param.data.icmph.code = code;
595 icmp_param.data.icmph.un.gateway = info;
596 icmp_param.data.icmph.checksum = 0;
597 icmp_param.skb = skb_in;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700598 icmp_param.offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800599 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 ipc.addr = iph->saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000601 ipc.opt = &icmp_param.replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000602 ipc.tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
David S. Miller77968b72011-05-08 17:12:19 -0700604 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos,
David S. Millerf6d460c2011-03-01 15:49:55 -0800605 type, code, &icmp_param);
606 if (IS_ERR(rt))
607 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800608
David S. Millera48eff12011-05-18 18:42:43 -0400609 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 goto ende;
611
612 /* RFC says return as much as we can without exceeding 576 bytes. */
613
Changli Gaod8d1f302010-06-10 23:31:35 -0700614 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 if (room > 576)
616 room = 576;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000617 room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 room -= sizeof(struct icmphdr);
619
620 icmp_param.data_len = skb_in->len - icmp_param.offset;
621 if (icmp_param.data_len > room)
622 icmp_param.data_len = room;
623 icmp_param.head_len = sizeof(struct icmphdr);
624
David S. Miller77968b72011-05-08 17:12:19 -0700625 icmp_push_reply(&icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626ende:
627 ip_rt_put(rt);
628out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800629 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630out:;
631}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000632EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
634
635/*
636 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
637 */
638
639static void icmp_unreach(struct sk_buff *skb)
640{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000641 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 struct icmphdr *icmph;
643 int hash, protocol;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000644 const struct net_protocol *ipprot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800646 struct net *net;
647
Eric Dumazetadf30902009-06-02 05:19:30 +0000648 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
650 /*
651 * Incomplete header ?
652 * Only checks for the IP header, there should be an
653 * additional check for longer headers in upper levels.
654 */
655
656 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
657 goto out_err;
658
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300659 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000660 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
662 if (iph->ihl < 5) /* Mangled header, drop. */
663 goto out_err;
664
665 if (icmph->type == ICMP_DEST_UNREACH) {
666 switch (icmph->code & 15) {
667 case ICMP_NET_UNREACH:
668 case ICMP_HOST_UNREACH:
669 case ICMP_PROT_UNREACH:
670 case ICMP_PORT_UNREACH:
671 break;
672 case ICMP_FRAG_NEEDED:
673 if (ipv4_config.no_pmtu_disc) {
Joe Perchesafd465032012-03-12 07:03:32 +0000674 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700675 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 } else {
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800677 info = ip_rt_frag_needed(net, iph,
Timo Teras0010e462008-04-29 03:32:25 -0700678 ntohs(icmph->un.frag.mtu),
679 skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 if (!info)
681 goto out;
682 }
683 break;
684 case ICMP_SR_FAILED:
Joe Perchesafd465032012-03-12 07:03:32 +0000685 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700686 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 break;
688 default:
689 break;
690 }
691 if (icmph->code > NR_ICMP_UNREACH)
692 goto out;
693 } else if (icmph->type == ICMP_PARAMETERPROB)
694 info = ntohl(icmph->un.gateway) >> 24;
695
696 /*
697 * Throw it at our lower layers
698 *
699 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
700 * header.
701 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
702 * transport layer.
703 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
704 * transport layer.
705 */
706
707 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300708 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 * bogus responses to broadcast frames. If you see this message
710 * first check your netmask matches at both ends, if it does then
711 * get the other vendor to fix their kit.
712 */
713
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700714 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800715 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 if (net_ratelimit())
Joe Perches058bd4d2012-03-11 18:36:11 +0000717 pr_warn("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
718 &ip_hdr(skb)->saddr,
719 icmph->type, icmph->code,
720 &iph->daddr, skb->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 goto out;
722 }
723
724 /* Checkin full IP header plus 8 bytes of protocol to
725 * avoid additional coding at protocol handlers.
726 */
727 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
728 goto out;
729
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000730 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 protocol = iph->protocol;
732
733 /*
734 * Deliver ICMP message to raw sockets. Pretty useless feature?
735 */
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800736 raw_icmp_error(skb, protocol, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 hash = protocol & (MAX_INET_PROTOS - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 rcu_read_lock();
740 ipprot = rcu_dereference(inet_protos[hash]);
741 if (ipprot && ipprot->err_handler)
742 ipprot->err_handler(skb, info);
743 rcu_read_unlock();
744
745out:
746 return;
747out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700748 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 goto out;
750}
751
752
753/*
754 * Handle ICMP_REDIRECT.
755 */
756
757static void icmp_redirect(struct sk_buff *skb)
758{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000759 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760
761 if (skb->len < sizeof(struct iphdr))
762 goto out_err;
763
764 /*
765 * Get the copied header of the packet that caused the redirect
766 */
767 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
768 goto out;
769
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000770 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300772 switch (icmp_hdr(skb)->code & 7) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 case ICMP_REDIR_NET:
774 case ICMP_REDIR_NETTOS:
775 /*
776 * As per RFC recommendations now handle it as a host redirect.
777 */
778 case ICMP_REDIR_HOST:
779 case ICMP_REDIR_HOSTTOS:
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700780 ip_rt_redirect(ip_hdr(skb)->saddr, iph->daddr,
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300781 icmp_hdr(skb)->un.gateway,
Ilia Sotnikovcef26852006-03-25 01:38:55 -0800782 iph->saddr, skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900784 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000785
786 /* Ping wants to see redirects.
787 * Let's pretend they are errors of sorts... */
788 if (iph->protocol == IPPROTO_ICMP &&
789 iph->ihl >= 5 &&
790 pskb_may_pull(skb, (iph->ihl<<2)+8)) {
791 ping_err(skb, icmp_hdr(skb)->un.gateway);
792 }
793
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794out:
795 return;
796out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700797 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 goto out;
799}
800
801/*
802 * Handle ICMP_ECHO ("ping") requests.
803 *
804 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
805 * requests.
806 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
807 * included in the reply.
808 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
809 * echo requests, MUST have default=NOT.
810 * See also WRT handling of options once they are done and working.
811 */
812
813static void icmp_echo(struct sk_buff *skb)
814{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700815 struct net *net;
816
Eric Dumazetadf30902009-06-02 05:19:30 +0000817 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700818 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 struct icmp_bxm icmp_param;
820
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300821 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
823 icmp_param.skb = skb;
824 icmp_param.offset = 0;
825 icmp_param.data_len = skb->len;
826 icmp_param.head_len = sizeof(struct icmphdr);
827 icmp_reply(&icmp_param, skb);
828 }
829}
830
831/*
832 * Handle ICMP Timestamp requests.
833 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
834 * SHOULD be in the kernel for minimum random latency.
835 * MUST be accurate to a few minutes.
836 * MUST be updated at least at 15Hz.
837 */
838static void icmp_timestamp(struct sk_buff *skb)
839{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700840 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 struct icmp_bxm icmp_param;
842 /*
843 * Too short.
844 */
845 if (skb->len < 4)
846 goto out_err;
847
848 /*
849 * Fill in the current time as ms since midnight UT:
850 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700851 getnstimeofday(&tv);
852 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
853 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 icmp_param.data.times[2] = icmp_param.data.times[1];
855 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
856 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300857 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
859 icmp_param.data.icmph.code = 0;
860 icmp_param.skb = skb;
861 icmp_param.offset = 0;
862 icmp_param.data_len = 0;
863 icmp_param.head_len = sizeof(struct icmphdr) + 12;
864 icmp_reply(&icmp_param, skb);
865out:
866 return;
867out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000868 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 goto out;
870}
871
872
873/*
874 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
875 *
876 * RFC1122 (3.2.2.9). A host MUST only send replies to
877 * ADDRESS_MASK requests if it's been configured as an address mask
878 * agent. Receiving a request doesn't constitute implicit permission to
879 * act as one. Of course, implementing this correctly requires (SHOULD)
880 * a way to turn the functionality on and off. Another one for sysctl(),
881 * I guess. -- MS
882 *
883 * RFC1812 (4.3.3.9). A router MUST implement it.
884 * A router SHOULD have switch turning it on/off.
885 * This switch MUST be ON by default.
886 *
887 * Gratuitous replies, zero-source replies are not implemented,
888 * that complies with RFC. DO NOT implement them!!! All the idea
889 * of broadcast addrmask replies as specified in RFC950 is broken.
890 * The problem is that it is not uncommon to have several prefixes
891 * on one physical interface. Moreover, addrmask agent can even be
892 * not aware of existing another prefixes.
893 * If source is zero, addrmask agent cannot choose correct prefix.
894 * Gratuitous mask announcements suffer from the same problem.
895 * RFC1812 explains it, but still allows to use ADDRMASK,
896 * that is pretty silly. --ANK
897 *
898 * All these rules are so bizarre, that I removed kernel addrmask
899 * support at all. It is wrong, it is obsolete, nobody uses it in
900 * any case. --ANK
901 *
902 * Furthermore you can do it with a usermode address agent program
903 * anyway...
904 */
905
906static void icmp_address(struct sk_buff *skb)
907{
908#if 0
909 if (net_ratelimit())
910 printk(KERN_DEBUG "a guy asks for address mask. Who is it?\n");
911#endif
912}
913
914/*
915 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
916 * loudly if an inconsistency is found.
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000917 * called with rcu_read_lock()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 */
919
920static void icmp_address_reply(struct sk_buff *skb)
921{
Eric Dumazet511c3f92009-06-02 05:14:27 +0000922 struct rtable *rt = skb_rtable(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 struct net_device *dev = skb->dev;
924 struct in_device *in_dev;
925 struct in_ifaddr *ifa;
926
927 if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC))
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000928 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000930 in_dev = __in_dev_get_rcu(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 if (!in_dev)
Eric Dumazetcfa087f2010-06-07 22:34:35 +0000932 return;
933
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 if (in_dev->ifa_list &&
935 IN_DEV_LOG_MARTIANS(in_dev) &&
936 IN_DEV_FORWARD(in_dev)) {
Al Viroa144ea42006-09-28 18:00:55 -0700937 __be32 _mask, *mp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938
939 mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask);
Kris Katterjohn09a62662006-01-08 22:24:28 -0800940 BUG_ON(mp == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
942 if (*mp == ifa->ifa_mask &&
David S. Miller9f6abb52011-05-09 13:28:22 -0700943 inet_ifa_match(ip_hdr(skb)->saddr, ifa))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 break;
945 }
946 if (!ifa && net_ratelimit()) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000947 pr_info("Wrong address mask %pI4 from %s/%pI4\n",
948 mp, dev->name, &ip_hdr(skb)->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 }
950 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951}
952
953static void icmp_discard(struct sk_buff *skb)
954{
955}
956
957/*
958 * Deal with incoming ICMP packets.
959 */
960int icmp_rcv(struct sk_buff *skb)
961{
962 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000963 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700964 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
Herbert Xuaebcf822007-12-12 18:54:16 -0800966 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700967 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800968 int nh;
969
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700970 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800971 XFRM_STATE_ICMP))
972 goto drop;
973
Herbert Xu8b7817f2007-12-12 10:44:43 -0800974 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
975 goto drop;
976
977 nh = skb_network_offset(skb);
978 skb_set_network_header(skb, sizeof(*icmph));
979
980 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
981 goto drop;
982
983 skb_set_network_header(skb, nh);
984 }
985
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700986 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
988 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700989 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800990 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 break;
Herbert Xufb286bb2005-11-10 13:01:24 -0800992 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -0800994 skb->csum = 0;
995 if (__skb_checksum_complete(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 }
998
Herbert Xu8cf22942008-02-05 03:15:50 -0800999 if (!pskb_pull(skb, sizeof(*icmph)))
1000 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -03001002 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Pavel Emelyanovf66ac032008-07-14 23:04:00 -07001004 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 /*
1006 * 18 is the highest 'known' ICMP type. Anything else is a mystery
1007 *
1008 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
1009 * discarded.
1010 */
1011 if (icmph->type > NR_ICMP_TYPES)
1012 goto error;
1013
1014
1015 /*
1016 * Parse the ICMP message
1017 */
1018
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001019 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 /*
1021 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
1022 * silently ignored (we let user decide with a sysctl).
1023 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1024 * discarded if to broadcast/multicast.
1025 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -07001026 if ((icmph->type == ICMP_ECHO ||
1027 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -07001028 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 goto error;
1030 }
1031 if (icmph->type != ICMP_ECHO &&
1032 icmph->type != ICMP_TIMESTAMP &&
1033 icmph->type != ICMP_ADDRESS &&
1034 icmph->type != ICMP_ADDRESSREPLY) {
1035 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001036 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 }
1038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 icmp_pointers[icmph->type].handler(skb);
1040
1041drop:
1042 kfree_skb(skb);
1043 return 0;
1044error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001045 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 goto drop;
1047}
1048
1049/*
1050 * This table is the definition of how we handle ICMP.
1051 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -08001052static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001054 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 },
1056 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 .handler = icmp_discard,
1058 .error = 1,
1059 },
1060 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 .handler = icmp_discard,
1062 .error = 1,
1063 },
1064 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 .handler = icmp_unreach,
1066 .error = 1,
1067 },
1068 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 .handler = icmp_unreach,
1070 .error = 1,
1071 },
1072 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 .handler = icmp_redirect,
1074 .error = 1,
1075 },
1076 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 .handler = icmp_discard,
1078 .error = 1,
1079 },
1080 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 .handler = icmp_discard,
1082 .error = 1,
1083 },
1084 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 .handler = icmp_echo,
1086 },
1087 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 .handler = icmp_discard,
1089 .error = 1,
1090 },
1091 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 .handler = icmp_discard,
1093 .error = 1,
1094 },
1095 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 .handler = icmp_unreach,
1097 .error = 1,
1098 },
1099 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 .handler = icmp_unreach,
1101 .error = 1,
1102 },
1103 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 .handler = icmp_timestamp,
1105 },
1106 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 .handler = icmp_discard,
1108 },
1109 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 .handler = icmp_discard,
1111 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001112 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 .handler = icmp_discard,
1114 },
1115 [ICMP_ADDRESS] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 .handler = icmp_address,
1117 },
1118 [ICMP_ADDRESSREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 .handler = icmp_address_reply,
1120 },
1121};
1122
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001123static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 int i;
1126
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001127 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001128 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001129 kfree(net->ipv4.icmp_sk);
1130 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001131}
1132
Adrian Bunk263173a2008-04-21 02:31:23 -07001133static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001134{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001135 int i, err;
1136
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001137 net->ipv4.icmp_sk =
1138 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1139 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001140 return -ENOMEM;
1141
Denis V. Luneva5710d62008-02-29 11:14:50 -08001142 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001143 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001145 err = inet_ctl_sock_create(&sk, PF_INET,
1146 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001148 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001150 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
1152 /* Enough space for 2 64K ICMP packets, including
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001153 * sk_buff/skb_shared_info struct overhead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 */
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001155 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001157 /*
1158 * Speedup sock_wfree()
1159 */
1160 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001161 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001163
1164 /* Control parameters for ECHO replies. */
1165 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1166 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1167
1168 /* Control parameter - ignore bogus broadcast responses? */
1169 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1170
1171 /*
1172 * Configurable global rate limit.
1173 *
1174 * ratelimit defines tokens/packet consumed for dst->rate_token
1175 * bucket ratemask defines which icmp types are ratelimited by
1176 * setting it's bit position.
1177 *
1178 * default:
1179 * dest unreachable (3), source quench (4),
1180 * time exceeded (11), parameter problem (12)
1181 */
1182
1183 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1184 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1185 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1186
Denis V. Luneva5710d62008-02-29 11:14:50 -08001187 return 0;
1188
1189fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001190 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001191 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001192 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001193 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194}
1195
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001196static struct pernet_operations __net_initdata icmp_sk_ops = {
1197 .init = icmp_sk_init,
1198 .exit = icmp_sk_exit,
1199};
1200
1201int __init icmp_init(void)
1202{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001203 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001204}