blob: 5882f584910edcd38affd3be200df9d0ddf1ee14 [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
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700234int sysctl_icmp_msgs_per_sec __read_mostly = 1000;
235int sysctl_icmp_msgs_burst __read_mostly = 50;
236
237static struct {
238 spinlock_t lock;
239 u32 credit;
240 u32 stamp;
241} icmp_global = {
242 .lock = __SPIN_LOCK_UNLOCKED(icmp_global.lock),
243};
244
245/**
246 * icmp_global_allow - Are we allowed to send one more ICMP message ?
247 *
248 * Uses a token bucket to limit our ICMP messages to sysctl_icmp_msgs_per_sec.
249 * Returns false if we reached the limit and can not send another packet.
250 * Note: called with BH disabled
251 */
252bool icmp_global_allow(void)
253{
254 u32 credit, delta, incr = 0, now = (u32)jiffies;
255 bool rc = false;
256
257 /* Check if token bucket is empty and cannot be refilled
258 * without taking the spinlock.
259 */
260 if (!icmp_global.credit) {
261 delta = min_t(u32, now - icmp_global.stamp, HZ);
262 if (delta < HZ / 50)
263 return false;
264 }
265
266 spin_lock(&icmp_global.lock);
267 delta = min_t(u32, now - icmp_global.stamp, HZ);
268 if (delta >= HZ / 50) {
269 incr = sysctl_icmp_msgs_per_sec * delta / HZ ;
270 if (incr)
271 icmp_global.stamp = now;
272 }
273 credit = min_t(u32, icmp_global.credit + incr, sysctl_icmp_msgs_burst);
274 if (credit) {
275 credit--;
276 rc = true;
277 }
278 icmp_global.credit = credit;
279 spin_unlock(&icmp_global.lock);
280 return rc;
281}
282EXPORT_SYMBOL(icmp_global_allow);
283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284/*
285 * Send an ICMP frame.
286 */
287
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700288static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
289 struct flowi4 *fl4, int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290{
Changli Gaod8d1f302010-06-10 23:31:35 -0700291 struct dst_entry *dst = &rt->dst;
David S. Miller92d86822011-02-04 15:55:25 -0800292 bool rc = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
294 if (type > NR_ICMP_TYPES)
295 goto out;
296
297 /* Don't limit PMTU discovery. */
298 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
299 goto out;
300
301 /* No rate limit on loopback */
302 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900303 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304
305 /* Limit if icmp type is enabled in ratemask. */
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700306 if (!((1 << type) & net->ipv4.sysctl_icmp_ratemask))
307 goto out;
308
309 rc = false;
310 if (icmp_global_allow()) {
311 struct inet_peer *peer;
312
313 peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, 1);
David S. Millerfbfe95a2012-06-08 23:24:18 -0700314 rc = inet_peer_xrlim_allow(peer,
David S. Miller92d86822011-02-04 15:55:25 -0800315 net->ipv4.sysctl_icmp_ratelimit);
Neal Cardwelle1a67642012-11-24 18:54:37 +0000316 if (peer)
317 inet_putpeer(peer);
David S. Miller92d86822011-02-04 15:55:25 -0800318 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319out:
320 return rc;
321}
322
323/*
324 * Maintain the counters used in the SNMP statistics for outgoing ICMP
325 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700326void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700328 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700329 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330}
331
332/*
333 * Checksum each fragment, and on the first include the headers and final
334 * checksum.
335 */
336static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
337 struct sk_buff *skb)
338{
339 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800340 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341
342 csum = skb_copy_and_csum_bits(icmp_param->skb,
343 icmp_param->offset + offset,
344 to, len, 0);
345
346 skb->csum = csum_block_add(skb->csum, csum, odd);
347 if (icmp_pointers[icmp_param->data.icmph.type].error)
348 nf_ct_attach(skb, icmp_param->skb);
349 return 0;
350}
351
352static void icmp_push_reply(struct icmp_bxm *icmp_param,
David S. Miller77968b72011-05-08 17:12:19 -0700353 struct flowi4 *fl4,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800354 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800356 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 struct sk_buff *skb;
358
Changli Gaod8d1f302010-06-10 23:31:35 -0700359 sk = icmp_sk(dev_net((*rt)->dst.dev));
David S. Millerf5fca602011-05-08 17:24:10 -0700360 if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900361 icmp_param->data_len+icmp_param->head_len,
362 icmp_param->head_len,
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700363 ipc, rt, MSG_DONTWAIT) < 0) {
364 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800365 ip_flush_pending_frames(sk);
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700366 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300367 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800368 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 struct sk_buff *skb1;
370
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800371 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 csum = csum_add(csum, skb1->csum);
373 }
374 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
375 (char *)icmph,
376 icmp_param->head_len, csum);
377 icmph->checksum = csum_fold(csum);
378 skb->ip_summed = CHECKSUM_NONE;
David S. Miller77968b72011-05-08 17:12:19 -0700379 ip_push_pending_frames(sk, fl4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 }
381}
382
383/*
384 * Driving logic for building and sending ICMP messages.
385 */
386
387static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
388{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000390 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700391 struct net *net = dev_net(rt->dst.dev);
David S. Miller77968b72011-05-08 17:12:19 -0700392 struct flowi4 fl4;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700393 struct sock *sk;
394 struct inet_sock *inet;
David S. Miller35ebf652012-06-28 03:59:11 -0700395 __be32 daddr, saddr;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700396 u32 mark = IP4_REPLY_MARK(net, skb->mark);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000398 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb))
Hormsf00c4012006-02-02 17:03:18 -0800399 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700401 sk = icmp_xmit_lock(net);
402 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700404 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
406 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700408 inet->tos = ip_hdr(skb)->tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700409 sk->sk_mark = mark;
David S. Miller9f6abb52011-05-09 13:28:22 -0700410 daddr = ipc.addr = ip_hdr(skb)->saddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700411 saddr = fib_compute_spec_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000413 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200414 ipc.ttl = 0;
415 ipc.tos = -1;
416
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000417 if (icmp_param->replyopts.opt.opt.optlen) {
418 ipc.opt = &icmp_param->replyopts.opt;
419 if (ipc.opt->opt.srr)
420 daddr = icmp_param->replyopts.opt.opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 }
David S. Miller77968b72011-05-08 17:12:19 -0700422 memset(&fl4, 0, sizeof(fl4));
423 fl4.daddr = daddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700424 fl4.saddr = saddr;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700425 fl4.flowi4_mark = mark;
David S. Miller77968b72011-05-08 17:12:19 -0700426 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
427 fl4.flowi4_proto = IPPROTO_ICMP;
428 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
429 rt = ip_route_output_key(net, &fl4);
430 if (IS_ERR(rt))
431 goto out_unlock;
David S. Millera48eff12011-05-18 18:42:43 -0400432 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 icmp_param->data.icmph.code))
David S. Miller77968b72011-05-08 17:12:19 -0700434 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 ip_rt_put(rt);
436out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800437 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438}
439
David S. Miller77968b72011-05-08 17:12:19 -0700440static struct rtable *icmp_route_lookup(struct net *net,
441 struct flowi4 *fl4,
442 struct sk_buff *skb_in,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000443 const struct iphdr *iph,
Lorenzo Colittie1108612014-05-13 10:17:33 -0700444 __be32 saddr, u8 tos, u32 mark,
David S. Millerf6d460c2011-03-01 15:49:55 -0800445 int type, int code,
446 struct icmp_bxm *param)
447{
David S. Millerf6d460c2011-03-01 15:49:55 -0800448 struct rtable *rt, *rt2;
David S. Miller415b3332011-07-22 06:22:10 -0700449 struct flowi4 fl4_dec;
David S. Millerf6d460c2011-03-01 15:49:55 -0800450 int err;
451
David S. Miller77968b72011-05-08 17:12:19 -0700452 memset(fl4, 0, sizeof(*fl4));
453 fl4->daddr = (param->replyopts.opt.opt.srr ?
454 param->replyopts.opt.opt.faddr : iph->saddr);
455 fl4->saddr = saddr;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700456 fl4->flowi4_mark = mark;
David S. Miller77968b72011-05-08 17:12:19 -0700457 fl4->flowi4_tos = RT_TOS(tos);
458 fl4->flowi4_proto = IPPROTO_ICMP;
459 fl4->fl4_icmp_type = type;
460 fl4->fl4_icmp_code = code;
461 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
462 rt = __ip_route_output_key(net, fl4);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800463 if (IS_ERR(rt))
464 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800465
466 /* No need to clone since we're just using its address. */
467 rt2 = rt;
468
David S. Miller9d6ec932011-03-12 01:12:47 -0500469 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
David S. Miller77968b72011-05-08 17:12:19 -0700470 flowi4_to_flowi(fl4), NULL, 0);
David S. Miller452edd52011-03-02 13:27:41 -0800471 if (!IS_ERR(rt)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800472 if (rt != rt2)
473 return rt;
David S. Miller452edd52011-03-02 13:27:41 -0800474 } else if (PTR_ERR(rt) == -EPERM) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800475 rt = NULL;
David S. Miller452edd52011-03-02 13:27:41 -0800476 } else
477 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800478
David S. Miller415b3332011-07-22 06:22:10 -0700479 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
David S. Millerf6d460c2011-03-01 15:49:55 -0800480 if (err)
481 goto relookup_failed;
482
David S. Miller415b3332011-07-22 06:22:10 -0700483 if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) {
484 rt2 = __ip_route_output_key(net, &fl4_dec);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800485 if (IS_ERR(rt2))
486 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800487 } else {
David S. Miller9d6ec932011-03-12 01:12:47 -0500488 struct flowi4 fl4_2 = {};
David S. Millerf6d460c2011-03-01 15:49:55 -0800489 unsigned long orefdst;
490
David S. Miller415b3332011-07-22 06:22:10 -0700491 fl4_2.daddr = fl4_dec.saddr;
David S. Miller9d6ec932011-03-12 01:12:47 -0500492 rt2 = ip_route_output_key(net, &fl4_2);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800493 if (IS_ERR(rt2)) {
494 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800495 goto relookup_failed;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800496 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800497 /* Ugh! */
498 orefdst = skb_in->_skb_refdst; /* save old refdst */
David S. Miller415b3332011-07-22 06:22:10 -0700499 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
David S. Millerf6d460c2011-03-01 15:49:55 -0800500 RT_TOS(tos), rt2->dst.dev);
501
502 dst_release(&rt2->dst);
503 rt2 = skb_rtable(skb_in);
504 skb_in->_skb_refdst = orefdst; /* restore old refdst */
505 }
506
507 if (err)
508 goto relookup_failed;
509
David S. Miller9d6ec932011-03-12 01:12:47 -0500510 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
David S. Miller415b3332011-07-22 06:22:10 -0700511 flowi4_to_flowi(&fl4_dec), NULL,
David S. Miller9d6ec932011-03-12 01:12:47 -0500512 XFRM_LOOKUP_ICMP);
David S. Miller452edd52011-03-02 13:27:41 -0800513 if (!IS_ERR(rt2)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800514 dst_release(&rt->dst);
David S. Miller415b3332011-07-22 06:22:10 -0700515 memcpy(fl4, &fl4_dec, sizeof(*fl4));
David S. Millerf6d460c2011-03-01 15:49:55 -0800516 rt = rt2;
David S. Miller452edd52011-03-02 13:27:41 -0800517 } else if (PTR_ERR(rt2) == -EPERM) {
518 if (rt)
519 dst_release(&rt->dst);
520 return rt2;
521 } else {
522 err = PTR_ERR(rt2);
523 goto relookup_failed;
David S. Millerf6d460c2011-03-01 15:49:55 -0800524 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800525 return rt;
526
527relookup_failed:
528 if (rt)
529 return rt;
530 return ERR_PTR(err);
531}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
533/*
534 * Send an ICMP message in response to a situation
535 *
536 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
537 * MAY send more (we do).
538 * MUST NOT change this header information.
539 * MUST NOT reply to a multicast/broadcast IP address.
540 * MUST NOT reply to a multicast/broadcast MAC address.
541 * MUST reply to only the first fragment.
542 */
543
Al Viroe4883012006-09-26 22:17:28 -0700544void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545{
546 struct iphdr *iph;
547 int room;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000548 struct icmp_bxm *icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000549 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 struct ipcm_cookie ipc;
David S. Miller77968b72011-05-08 17:12:19 -0700551 struct flowi4 fl4;
Al Viroa61ced52006-09-26 21:27:54 -0700552 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 u8 tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700554 u32 mark;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800555 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800556 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557
558 if (!rt)
559 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700560 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561
562 /*
563 * Find the original header. It is expected to be valid, of course.
564 * Check this, icmp_send is called from the most obscure devices
565 * sometimes.
566 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700567 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700569 if ((u8 *)iph < skb_in->head ||
Simon Hormanf7c0c2a2013-05-28 20:34:27 +0000570 (skb_network_header(skb_in) + sizeof(*iph)) >
571 skb_tail_pointer(skb_in))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 goto out;
573
574 /*
575 * No replies to physical multicast/broadcast
576 */
577 if (skb_in->pkt_type != PACKET_HOST)
578 goto out;
579
580 /*
581 * Now check at the protocol level
582 */
583 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
584 goto out;
585
586 /*
587 * Only reply to fragment 0. We byte re-order the constant
588 * mask for efficiency.
589 */
590 if (iph->frag_off & htons(IP_OFFSET))
591 goto out;
592
593 /*
594 * If we send an ICMP error to an ICMP error a mess would result..
595 */
596 if (icmp_pointers[type].error) {
597 /*
598 * We are an error, check if we are replying to an
599 * ICMP error
600 */
601 if (iph->protocol == IPPROTO_ICMP) {
602 u8 _inner_type, *itp;
603
604 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700605 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 (iph->ihl << 2) +
607 offsetof(struct icmphdr,
608 type) -
609 skb_in->data,
610 sizeof(_inner_type),
611 &_inner_type);
612 if (itp == NULL)
613 goto out;
614
615 /*
616 * Assume any unknown ICMP type is an error. This
617 * isn't specified by the RFC, but think about it..
618 */
619 if (*itp > NR_ICMP_TYPES ||
620 icmp_pointers[*itp].error)
621 goto out;
622 }
623 }
624
Cong Wang9a99d4a2013-06-02 15:00:52 +0000625 icmp_param = kmalloc(sizeof(*icmp_param), GFP_ATOMIC);
626 if (!icmp_param)
627 return;
628
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700629 sk = icmp_xmit_lock(net);
630 if (sk == NULL)
Cong Wang9a99d4a2013-06-02 15:00:52 +0000631 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632
633 /*
634 * Construct source address and options.
635 */
636
637 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700638 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700639 struct net_device *dev = NULL;
640
Eric Dumazet685c7942009-11-01 19:31:03 +0000641 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800642 if (rt_is_input_route(rt) &&
643 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller92101b32012-07-23 16:29:00 -0700644 dev = dev_get_by_index_rcu(net, inet_iif(skb_in));
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700645
Eric Dumazet685c7942009-11-01 19:31:03 +0000646 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700647 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000648 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700649 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000650 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700651 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
653 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
654 IPTOS_PREC_INTERNETCONTROL) :
655 iph->tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700656 mark = IP4_REPLY_MARK(net, skb_in->mark);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Cong Wang9a99d4a2013-06-02 15:00:52 +0000658 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800659 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
661
662 /*
663 * Prepare data for ICMP header.
664 */
665
Cong Wang9a99d4a2013-06-02 15:00:52 +0000666 icmp_param->data.icmph.type = type;
667 icmp_param->data.icmph.code = code;
668 icmp_param->data.icmph.un.gateway = info;
669 icmp_param->data.icmph.checksum = 0;
670 icmp_param->skb = skb_in;
671 icmp_param->offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800672 inet_sk(sk)->tos = tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700673 sk->sk_mark = mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 ipc.addr = iph->saddr;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000675 ipc.opt = &icmp_param->replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000676 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200677 ipc.ttl = 0;
678 ipc.tos = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
Lorenzo Colittie1108612014-05-13 10:17:33 -0700680 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, mark,
Cong Wang9a99d4a2013-06-02 15:00:52 +0000681 type, code, icmp_param);
David S. Millerf6d460c2011-03-01 15:49:55 -0800682 if (IS_ERR(rt))
683 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800684
David S. Millera48eff12011-05-18 18:42:43 -0400685 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 goto ende;
687
688 /* RFC says return as much as we can without exceeding 576 bytes. */
689
Changli Gaod8d1f302010-06-10 23:31:35 -0700690 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 if (room > 576)
692 room = 576;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000693 room -= sizeof(struct iphdr) + icmp_param->replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 room -= sizeof(struct icmphdr);
695
Cong Wang9a99d4a2013-06-02 15:00:52 +0000696 icmp_param->data_len = skb_in->len - icmp_param->offset;
697 if (icmp_param->data_len > room)
698 icmp_param->data_len = room;
699 icmp_param->head_len = sizeof(struct icmphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
Cong Wang9a99d4a2013-06-02 15:00:52 +0000701 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702ende:
703 ip_rt_put(rt);
704out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800705 icmp_xmit_unlock(sk);
Cong Wang9a99d4a2013-06-02 15:00:52 +0000706out_free:
707 kfree(icmp_param);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708out:;
709}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000710EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
712
David S. Miller1de92432012-07-11 18:32:17 -0700713static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
714{
715 const struct iphdr *iph = (const struct iphdr *) skb->data;
716 const struct net_protocol *ipprot;
717 int protocol = iph->protocol;
718
David S. Millerf0a70e92012-07-12 08:06:04 -0700719 /* Checkin full IP header plus 8 bytes of protocol to
720 * avoid additional coding at protocol handlers.
721 */
Duan Jiong7304fe42014-07-31 17:54:32 +0800722 if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) {
723 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
David S. Millerf0a70e92012-07-12 08:06:04 -0700724 return;
Duan Jiong7304fe42014-07-31 17:54:32 +0800725 }
David S. Millerf0a70e92012-07-12 08:06:04 -0700726
David S. Miller1de92432012-07-11 18:32:17 -0700727 raw_icmp_error(skb, protocol, info);
728
David S. Miller1de92432012-07-11 18:32:17 -0700729 ipprot = rcu_dereference(inet_protos[protocol]);
730 if (ipprot && ipprot->err_handler)
731 ipprot->err_handler(skb, info);
David S. Miller1de92432012-07-11 18:32:17 -0700732}
733
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +0100734static bool icmp_tag_validation(int proto)
735{
736 bool ok;
737
738 rcu_read_lock();
739 ok = rcu_dereference(inet_protos[proto])->icmp_strict_tag_validation;
740 rcu_read_unlock();
741 return ok;
742}
743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744/*
Rami Rosen08578d82013-06-03 00:23:25 +0000745 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, ICMP_QUENCH, and
746 * ICMP_PARAMETERPROB.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 */
748
749static void icmp_unreach(struct sk_buff *skb)
750{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000751 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 struct icmphdr *icmph;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800753 struct net *net;
David S. Millerf9242b62012-06-19 18:56:21 -0700754 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800755
Eric Dumazetadf30902009-06-02 05:19:30 +0000756 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
758 /*
759 * Incomplete header ?
760 * Only checks for the IP header, there should be an
761 * additional check for longer headers in upper levels.
762 */
763
764 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
765 goto out_err;
766
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300767 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000768 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
770 if (iph->ihl < 5) /* Mangled header, drop. */
771 goto out_err;
772
773 if (icmph->type == ICMP_DEST_UNREACH) {
774 switch (icmph->code & 15) {
775 case ICMP_NET_UNREACH:
776 case ICMP_HOST_UNREACH:
777 case ICMP_PROT_UNREACH:
778 case ICMP_PORT_UNREACH:
779 break;
780 case ICMP_FRAG_NEEDED:
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +0100781 /* for documentation of the ip_no_pmtu_disc
782 * values please see
783 * Documentation/networking/ip-sysctl.txt
784 */
785 switch (net->ipv4.sysctl_ip_no_pmtu_disc) {
786 default:
Joe Perchesafd465032012-03-12 07:03:32 +0000787 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700788 &iph->daddr);
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +0100789 break;
790 case 2:
791 goto out;
792 case 3:
793 if (!icmp_tag_validation(iph->protocol))
794 goto out;
795 /* fall through */
796 case 0:
David S. Miller46517002012-06-10 00:04:12 -0700797 info = ntohs(icmph->un.frag.mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 }
799 break;
800 case ICMP_SR_FAILED:
Joe Perchesafd465032012-03-12 07:03:32 +0000801 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700802 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 break;
804 default:
805 break;
806 }
807 if (icmph->code > NR_ICMP_UNREACH)
808 goto out;
809 } else if (icmph->type == ICMP_PARAMETERPROB)
810 info = ntohl(icmph->un.gateway) >> 24;
811
812 /*
813 * Throw it at our lower layers
814 *
815 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
816 * header.
817 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
818 * transport layer.
819 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
820 * transport layer.
821 */
822
823 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300824 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 * bogus responses to broadcast frames. If you see this message
826 * first check your netmask matches at both ends, if it does then
827 * get the other vendor to fix their kit.
828 */
829
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700830 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800831 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Joe Perchese87cc472012-05-13 21:56:26 +0000832 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
833 &ip_hdr(skb)->saddr,
834 icmph->type, icmph->code,
835 &iph->daddr, skb->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 goto out;
837 }
838
David S. Miller1de92432012-07-11 18:32:17 -0700839 icmp_socket_deliver(skb, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840
841out:
842 return;
843out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700844 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 goto out;
846}
847
848
849/*
850 * Handle ICMP_REDIRECT.
851 */
852
853static void icmp_redirect(struct sk_buff *skb)
854{
David S. Miller94206122012-07-11 20:38:08 -0700855 if (skb->len < sizeof(struct iphdr)) {
856 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
857 return;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900858 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000859
David S. Miller94206122012-07-11 20:38:08 -0700860 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
861 return;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000862
David S. Miller94206122012-07-11 20:38:08 -0700863 icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864}
865
866/*
867 * Handle ICMP_ECHO ("ping") requests.
868 *
869 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
870 * requests.
871 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
872 * included in the reply.
873 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
874 * echo requests, MUST have default=NOT.
875 * See also WRT handling of options once they are done and working.
876 */
877
878static void icmp_echo(struct sk_buff *skb)
879{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700880 struct net *net;
881
Eric Dumazetadf30902009-06-02 05:19:30 +0000882 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700883 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 struct icmp_bxm icmp_param;
885
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300886 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
888 icmp_param.skb = skb;
889 icmp_param.offset = 0;
890 icmp_param.data_len = skb->len;
891 icmp_param.head_len = sizeof(struct icmphdr);
892 icmp_reply(&icmp_param, skb);
893 }
894}
895
896/*
897 * Handle ICMP Timestamp requests.
898 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
899 * SHOULD be in the kernel for minimum random latency.
900 * MUST be accurate to a few minutes.
901 * MUST be updated at least at 15Hz.
902 */
903static void icmp_timestamp(struct sk_buff *skb)
904{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700905 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 struct icmp_bxm icmp_param;
907 /*
908 * Too short.
909 */
910 if (skb->len < 4)
911 goto out_err;
912
913 /*
914 * Fill in the current time as ms since midnight UT:
915 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700916 getnstimeofday(&tv);
917 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
918 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 icmp_param.data.times[2] = icmp_param.data.times[1];
920 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
921 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300922 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
924 icmp_param.data.icmph.code = 0;
925 icmp_param.skb = skb;
926 icmp_param.offset = 0;
927 icmp_param.data_len = 0;
928 icmp_param.head_len = sizeof(struct icmphdr) + 12;
929 icmp_reply(&icmp_param, skb);
930out:
931 return;
932out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000933 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 goto out;
935}
936
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937static void icmp_discard(struct sk_buff *skb)
938{
939}
940
941/*
942 * Deal with incoming ICMP packets.
943 */
944int icmp_rcv(struct sk_buff *skb)
945{
946 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000947 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700948 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
Herbert Xuaebcf822007-12-12 18:54:16 -0800950 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700951 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800952 int nh;
953
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700954 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800955 XFRM_STATE_ICMP))
956 goto drop;
957
Herbert Xu8b7817f2007-12-12 10:44:43 -0800958 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
959 goto drop;
960
961 nh = skb_network_offset(skb);
962 skb_set_network_header(skb, sizeof(*icmph));
963
964 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
965 goto drop;
966
967 skb_set_network_header(skb, nh);
968 }
969
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700970 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971
Tom Herbert29a96e12014-05-07 16:52:21 -0700972 if (skb_checksum_simple_validate(skb))
973 goto csum_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974
Herbert Xu8cf22942008-02-05 03:15:50 -0800975 if (!pskb_pull(skb, sizeof(*icmph)))
976 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300978 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979
Pavel Emelyanovf66ac032008-07-14 23:04:00 -0700980 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 /*
982 * 18 is the highest 'known' ICMP type. Anything else is a mystery
983 *
984 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
985 * discarded.
986 */
987 if (icmph->type > NR_ICMP_TYPES)
988 goto error;
989
990
991 /*
992 * Parse the ICMP message
993 */
994
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900995 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 /*
997 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
998 * silently ignored (we let user decide with a sysctl).
999 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1000 * discarded if to broadcast/multicast.
1001 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -07001002 if ((icmph->type == ICMP_ECHO ||
1003 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -07001004 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 goto error;
1006 }
1007 if (icmph->type != ICMP_ECHO &&
1008 icmph->type != ICMP_TIMESTAMP &&
1009 icmph->type != ICMP_ADDRESS &&
1010 icmph->type != ICMP_ADDRESSREPLY) {
1011 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001012 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 }
1014
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 icmp_pointers[icmph->type].handler(skb);
1016
1017drop:
1018 kfree_skb(skb);
1019 return 0;
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001020csum_error:
1021 ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001023 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 goto drop;
1025}
1026
Li Wei5b052042013-02-21 22:18:44 +00001027void icmp_err(struct sk_buff *skb, u32 info)
1028{
1029 struct iphdr *iph = (struct iphdr *)skb->data;
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +00001030 int offset = iph->ihl<<2;
1031 struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset);
Li Wei5b052042013-02-21 22:18:44 +00001032 int type = icmp_hdr(skb)->type;
1033 int code = icmp_hdr(skb)->code;
1034 struct net *net = dev_net(skb->dev);
1035
1036 /*
1037 * Use ping_err to handle all icmp errors except those
1038 * triggered by ICMP_ECHOREPLY which sent from kernel.
1039 */
1040 if (icmph->type != ICMP_ECHOREPLY) {
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +00001041 ping_err(skb, offset, info);
Li Wei5b052042013-02-21 22:18:44 +00001042 return;
1043 }
1044
1045 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
1046 ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0);
1047 else if (type == ICMP_REDIRECT)
1048 ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0);
1049}
1050
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051/*
1052 * This table is the definition of how we handle ICMP.
1053 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -08001054static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001056 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 },
1058 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 .handler = icmp_discard,
1060 .error = 1,
1061 },
1062 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 .handler = icmp_discard,
1064 .error = 1,
1065 },
1066 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 .handler = icmp_unreach,
1068 .error = 1,
1069 },
1070 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 .handler = icmp_unreach,
1072 .error = 1,
1073 },
1074 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 .handler = icmp_redirect,
1076 .error = 1,
1077 },
1078 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 .handler = icmp_discard,
1080 .error = 1,
1081 },
1082 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 .handler = icmp_discard,
1084 .error = 1,
1085 },
1086 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 .handler = icmp_echo,
1088 },
1089 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 .handler = icmp_discard,
1091 .error = 1,
1092 },
1093 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 .handler = icmp_discard,
1095 .error = 1,
1096 },
1097 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 .handler = icmp_unreach,
1099 .error = 1,
1100 },
1101 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 .handler = icmp_unreach,
1103 .error = 1,
1104 },
1105 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 .handler = icmp_timestamp,
1107 },
1108 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 .handler = icmp_discard,
1110 },
1111 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 .handler = icmp_discard,
1113 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001114 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 .handler = icmp_discard,
1116 },
1117 [ICMP_ADDRESS] = {
David S. Miller838942a2012-07-23 13:20:26 -07001118 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 },
1120 [ICMP_ADDRESSREPLY] = {
David S. Miller838942a2012-07-23 13:20:26 -07001121 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 },
1123};
1124
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001125static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 int i;
1128
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001129 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001130 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001131 kfree(net->ipv4.icmp_sk);
1132 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001133}
1134
Adrian Bunk263173a2008-04-21 02:31:23 -07001135static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001136{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001137 int i, err;
1138
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001139 net->ipv4.icmp_sk =
1140 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1141 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001142 return -ENOMEM;
1143
Denis V. Luneva5710d62008-02-29 11:14:50 -08001144 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001145 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001147 err = inet_ctl_sock_create(&sk, PF_INET,
1148 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001150 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001152 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
1154 /* Enough space for 2 64K ICMP packets, including
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001155 * sk_buff/skb_shared_info struct overhead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156 */
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001157 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001159 /*
1160 * Speedup sock_wfree()
1161 */
1162 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001163 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001165
1166 /* Control parameters for ECHO replies. */
1167 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1168 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1169
1170 /* Control parameter - ignore bogus broadcast responses? */
1171 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1172
1173 /*
1174 * Configurable global rate limit.
1175 *
1176 * ratelimit defines tokens/packet consumed for dst->rate_token
1177 * bucket ratemask defines which icmp types are ratelimited by
1178 * setting it's bit position.
1179 *
1180 * default:
1181 * dest unreachable (3), source quench (4),
1182 * time exceeded (11), parameter problem (12)
1183 */
1184
1185 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1186 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1187 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1188
Denis V. Luneva5710d62008-02-29 11:14:50 -08001189 return 0;
1190
1191fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001192 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001193 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001194 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001195 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196}
1197
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001198static struct pernet_operations __net_initdata icmp_sk_ops = {
1199 .init = icmp_sk_init,
1200 .exit = icmp_sk_exit,
1201};
1202
1203int __init icmp_init(void)
1204{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001205 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001206}