blob: 36e26977c9088c1dbd09cd13e9a5e2c43369fe31 [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>
David Ahern385add92015-09-29 20:07:13 -070099#include <net/l3mdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
101/*
102 * Build xmit assembly blocks
103 */
104
105struct icmp_bxm {
106 struct sk_buff *skb;
107 int offset;
108 int data_len;
109
110 struct {
111 struct icmphdr icmph;
Al Virob03d73e2006-09-27 18:38:30 -0700112 __be32 times[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113 } data;
114 int head_len;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000115 struct ip_options_data replyopts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116};
117
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118/* An array of errno for error messages from dest unreach. */
119/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
120
Alexey Dobriyane7548342010-01-22 10:18:25 +0000121const struct icmp_err icmp_err_convert[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 {
123 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
124 .fatal = 0,
125 },
126 {
127 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
128 .fatal = 0,
129 },
130 {
131 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
132 .fatal = 1,
133 },
134 {
135 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
136 .fatal = 1,
137 },
138 {
139 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
140 .fatal = 0,
141 },
142 {
143 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
144 .fatal = 0,
145 },
146 {
147 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
148 .fatal = 1,
149 },
150 {
151 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
152 .fatal = 1,
153 },
154 {
155 .errno = ENONET, /* ICMP_HOST_ISOLATED */
156 .fatal = 1,
157 },
158 {
159 .errno = ENETUNREACH, /* ICMP_NET_ANO */
160 .fatal = 1,
161 },
162 {
163 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
164 .fatal = 1,
165 },
166 {
167 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
168 .fatal = 0,
169 },
170 {
171 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
172 .fatal = 0,
173 },
174 {
175 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
176 .fatal = 1,
177 },
178 {
179 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
180 .fatal = 1,
181 },
182 {
183 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
184 .fatal = 1,
185 },
186};
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000187EXPORT_SYMBOL(icmp_err_convert);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189/*
190 * ICMP control array. This specifies what to do with each ICMP.
191 */
192
193struct icmp_control {
Rick Jonese3e32172014-11-17 14:04:29 -0800194 bool (*handler)(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 short error; /* This ICMP is classed as an error message */
196};
197
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800198static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
200/*
201 * The ICMP socket(s). This is the most convenient way to flow control
202 * our ICMP output as well as maintain a clean interface throughout
203 * all layers. All Socketless IP sends will soon be gone.
204 *
205 * On SMP we have one ICMP socket per-cpu.
206 */
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800207static struct sock *icmp_sk(struct net *net)
208{
Eric Dumazet349c9e32015-01-29 15:58:09 -0800209 return *this_cpu_ptr(net->ipv4.icmp_sk);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800210}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700212static inline struct sock *icmp_xmit_lock(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213{
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700214 struct sock *sk;
215
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 local_bh_disable();
217
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700218 sk = icmp_sk(net);
219
Denis V. Lunev405666d2008-02-29 11:16:46 -0800220 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 /* This can happen if the output path signals a
222 * dst_link_failure() for an outgoing ICMP packet.
223 */
224 local_bh_enable();
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700225 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 }
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700227 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228}
229
Denis V. Lunev405666d2008-02-29 11:16:46 -0800230static inline void icmp_xmit_unlock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231{
Denis V. Lunev405666d2008-02-29 11:16:46 -0800232 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233}
234
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700235int sysctl_icmp_msgs_per_sec __read_mostly = 1000;
236int sysctl_icmp_msgs_burst __read_mostly = 50;
237
238static struct {
239 spinlock_t lock;
240 u32 credit;
241 u32 stamp;
242} icmp_global = {
243 .lock = __SPIN_LOCK_UNLOCKED(icmp_global.lock),
244};
245
246/**
247 * icmp_global_allow - Are we allowed to send one more ICMP message ?
248 *
249 * Uses a token bucket to limit our ICMP messages to sysctl_icmp_msgs_per_sec.
250 * Returns false if we reached the limit and can not send another packet.
251 * Note: called with BH disabled
252 */
253bool icmp_global_allow(void)
254{
255 u32 credit, delta, incr = 0, now = (u32)jiffies;
256 bool rc = false;
257
258 /* Check if token bucket is empty and cannot be refilled
259 * without taking the spinlock.
260 */
261 if (!icmp_global.credit) {
262 delta = min_t(u32, now - icmp_global.stamp, HZ);
263 if (delta < HZ / 50)
264 return false;
265 }
266
267 spin_lock(&icmp_global.lock);
268 delta = min_t(u32, now - icmp_global.stamp, HZ);
269 if (delta >= HZ / 50) {
270 incr = sysctl_icmp_msgs_per_sec * delta / HZ ;
271 if (incr)
272 icmp_global.stamp = now;
273 }
274 credit = min_t(u32, icmp_global.credit + incr, sysctl_icmp_msgs_burst);
275 if (credit) {
276 credit--;
277 rc = true;
278 }
279 icmp_global.credit = credit;
280 spin_unlock(&icmp_global.lock);
281 return rc;
282}
283EXPORT_SYMBOL(icmp_global_allow);
284
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285/*
286 * Send an ICMP frame.
287 */
288
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700289static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
290 struct flowi4 *fl4, int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291{
Changli Gaod8d1f302010-06-10 23:31:35 -0700292 struct dst_entry *dst = &rt->dst;
David S. Miller92d86822011-02-04 15:55:25 -0800293 bool rc = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294
295 if (type > NR_ICMP_TYPES)
296 goto out;
297
298 /* Don't limit PMTU discovery. */
299 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
300 goto out;
301
302 /* No rate limit on loopback */
303 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900304 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
306 /* Limit if icmp type is enabled in ratemask. */
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700307 if (!((1 << type) & net->ipv4.sysctl_icmp_ratemask))
308 goto out;
309
310 rc = false;
311 if (icmp_global_allow()) {
David Ahern385add92015-09-29 20:07:13 -0700312 int vif = l3mdev_master_ifindex(dst->dev);
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700313 struct inet_peer *peer;
314
David Ahern192132b2015-08-27 16:07:03 -0700315 peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, vif, 1);
David S. Millerfbfe95a2012-06-08 23:24:18 -0700316 rc = inet_peer_xrlim_allow(peer,
David S. Miller92d86822011-02-04 15:55:25 -0800317 net->ipv4.sysctl_icmp_ratelimit);
Neal Cardwelle1a67642012-11-24 18:54:37 +0000318 if (peer)
319 inet_putpeer(peer);
David S. Miller92d86822011-02-04 15:55:25 -0800320 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321out:
322 return rc;
323}
324
325/*
326 * Maintain the counters used in the SNMP statistics for outgoing ICMP
327 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700328void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700330 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700331 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332}
333
334/*
335 * Checksum each fragment, and on the first include the headers and final
336 * checksum.
337 */
338static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
339 struct sk_buff *skb)
340{
341 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800342 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
344 csum = skb_copy_and_csum_bits(icmp_param->skb,
345 icmp_param->offset + offset,
346 to, len, 0);
347
348 skb->csum = csum_block_add(skb->csum, csum, odd);
349 if (icmp_pointers[icmp_param->data.icmph.type].error)
350 nf_ct_attach(skb, icmp_param->skb);
351 return 0;
352}
353
354static void icmp_push_reply(struct icmp_bxm *icmp_param,
David S. Miller77968b72011-05-08 17:12:19 -0700355 struct flowi4 *fl4,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800356 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800358 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 struct sk_buff *skb;
360
Changli Gaod8d1f302010-06-10 23:31:35 -0700361 sk = icmp_sk(dev_net((*rt)->dst.dev));
David S. Millerf5fca602011-05-08 17:24:10 -0700362 if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900363 icmp_param->data_len+icmp_param->head_len,
364 icmp_param->head_len,
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700365 ipc, rt, MSG_DONTWAIT) < 0) {
366 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800367 ip_flush_pending_frames(sk);
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700368 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300369 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800370 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 struct sk_buff *skb1;
372
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800373 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 csum = csum_add(csum, skb1->csum);
375 }
376 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
377 (char *)icmph,
378 icmp_param->head_len, csum);
379 icmph->checksum = csum_fold(csum);
380 skb->ip_summed = CHECKSUM_NONE;
David S. Miller77968b72011-05-08 17:12:19 -0700381 ip_push_pending_frames(sk, fl4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 }
383}
384
385/*
386 * Driving logic for building and sending ICMP messages.
387 */
388
389static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
390{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000392 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700393 struct net *net = dev_net(rt->dst.dev);
David S. Miller77968b72011-05-08 17:12:19 -0700394 struct flowi4 fl4;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700395 struct sock *sk;
396 struct inet_sock *inet;
David S. Miller35ebf652012-06-28 03:59:11 -0700397 __be32 daddr, saddr;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700398 u32 mark = IP4_REPLY_MARK(net, skb->mark);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000400 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb))
Hormsf00c4012006-02-02 17:03:18 -0800401 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700403 sk = icmp_xmit_lock(net);
Ian Morris51456b22015-04-03 09:17:26 +0100404 if (!sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700406 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
408 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700410 inet->tos = ip_hdr(skb)->tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700411 sk->sk_mark = mark;
David S. Miller9f6abb52011-05-09 13:28:22 -0700412 daddr = ipc.addr = ip_hdr(skb)->saddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700413 saddr = fib_compute_spec_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000415 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200416 ipc.ttl = 0;
417 ipc.tos = -1;
418
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000419 if (icmp_param->replyopts.opt.opt.optlen) {
420 ipc.opt = &icmp_param->replyopts.opt;
421 if (ipc.opt->opt.srr)
422 daddr = icmp_param->replyopts.opt.opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 }
David S. Miller77968b72011-05-08 17:12:19 -0700424 memset(&fl4, 0, sizeof(fl4));
425 fl4.daddr = daddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700426 fl4.saddr = saddr;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700427 fl4.flowi4_mark = mark;
David S. Miller77968b72011-05-08 17:12:19 -0700428 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
429 fl4.flowi4_proto = IPPROTO_ICMP;
David Ahern385add92015-09-29 20:07:13 -0700430 fl4.flowi4_oif = l3mdev_master_ifindex(skb->dev);
David S. Miller77968b72011-05-08 17:12:19 -0700431 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
432 rt = ip_route_output_key(net, &fl4);
433 if (IS_ERR(rt))
434 goto out_unlock;
David S. Millera48eff12011-05-18 18:42:43 -0400435 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 icmp_param->data.icmph.code))
David S. Miller77968b72011-05-08 17:12:19 -0700437 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 ip_rt_put(rt);
439out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800440 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441}
442
Peter Nørlund79a13152015-09-30 10:12:22 +0200443#ifdef CONFIG_IP_ROUTE_MULTIPATH
444
445/* Source and destination is swapped. See ip_multipath_icmp_hash */
446static int icmp_multipath_hash_skb(const struct sk_buff *skb)
447{
448 const struct iphdr *iph = ip_hdr(skb);
449
450 return fib_multipath_hash(iph->daddr, iph->saddr);
451}
452
453#else
454
455#define icmp_multipath_hash_skb(skb) (-1)
456
457#endif
458
David S. Miller77968b72011-05-08 17:12:19 -0700459static struct rtable *icmp_route_lookup(struct net *net,
460 struct flowi4 *fl4,
461 struct sk_buff *skb_in,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000462 const struct iphdr *iph,
Lorenzo Colittie1108612014-05-13 10:17:33 -0700463 __be32 saddr, u8 tos, u32 mark,
David S. Millerf6d460c2011-03-01 15:49:55 -0800464 int type, int code,
465 struct icmp_bxm *param)
466{
David S. Millerf6d460c2011-03-01 15:49:55 -0800467 struct rtable *rt, *rt2;
David S. Miller415b3332011-07-22 06:22:10 -0700468 struct flowi4 fl4_dec;
David S. Millerf6d460c2011-03-01 15:49:55 -0800469 int err;
470
David S. Miller77968b72011-05-08 17:12:19 -0700471 memset(fl4, 0, sizeof(*fl4));
472 fl4->daddr = (param->replyopts.opt.opt.srr ?
473 param->replyopts.opt.opt.faddr : iph->saddr);
474 fl4->saddr = saddr;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700475 fl4->flowi4_mark = mark;
David S. Miller77968b72011-05-08 17:12:19 -0700476 fl4->flowi4_tos = RT_TOS(tos);
477 fl4->flowi4_proto = IPPROTO_ICMP;
478 fl4->fl4_icmp_type = type;
479 fl4->fl4_icmp_code = code;
David Ahern385add92015-09-29 20:07:13 -0700480 fl4->flowi4_oif = l3mdev_master_ifindex(skb_in->dev);
David Ahern613d09b2015-08-13 14:59:02 -0600481
David S. Miller77968b72011-05-08 17:12:19 -0700482 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
Peter Nørlund79a13152015-09-30 10:12:22 +0200483 rt = __ip_route_output_key_hash(net, fl4,
484 icmp_multipath_hash_skb(skb_in));
David S. Millerb23dd4f2011-03-02 14:31:35 -0800485 if (IS_ERR(rt))
486 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800487
488 /* No need to clone since we're just using its address. */
489 rt2 = rt;
490
David S. Miller9d6ec932011-03-12 01:12:47 -0500491 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
David S. Miller77968b72011-05-08 17:12:19 -0700492 flowi4_to_flowi(fl4), NULL, 0);
David S. Miller452edd52011-03-02 13:27:41 -0800493 if (!IS_ERR(rt)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800494 if (rt != rt2)
495 return rt;
David S. Miller452edd52011-03-02 13:27:41 -0800496 } else if (PTR_ERR(rt) == -EPERM) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800497 rt = NULL;
David S. Miller452edd52011-03-02 13:27:41 -0800498 } else
499 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800500
David S. Miller415b3332011-07-22 06:22:10 -0700501 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
David S. Millerf6d460c2011-03-01 15:49:55 -0800502 if (err)
503 goto relookup_failed;
504
David Ahern30bbaa12015-08-13 14:59:05 -0600505 if (inet_addr_type_dev_table(net, skb_in->dev,
506 fl4_dec.saddr) == RTN_LOCAL) {
David S. Miller415b3332011-07-22 06:22:10 -0700507 rt2 = __ip_route_output_key(net, &fl4_dec);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800508 if (IS_ERR(rt2))
509 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800510 } else {
David S. Miller9d6ec932011-03-12 01:12:47 -0500511 struct flowi4 fl4_2 = {};
David S. Millerf6d460c2011-03-01 15:49:55 -0800512 unsigned long orefdst;
513
David S. Miller415b3332011-07-22 06:22:10 -0700514 fl4_2.daddr = fl4_dec.saddr;
David S. Miller9d6ec932011-03-12 01:12:47 -0500515 rt2 = ip_route_output_key(net, &fl4_2);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800516 if (IS_ERR(rt2)) {
517 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800518 goto relookup_failed;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800519 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800520 /* Ugh! */
521 orefdst = skb_in->_skb_refdst; /* save old refdst */
Thomas Graf773a69d2015-07-21 10:43:55 +0200522 skb_dst_set(skb_in, NULL);
David S. Miller415b3332011-07-22 06:22:10 -0700523 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
David S. Millerf6d460c2011-03-01 15:49:55 -0800524 RT_TOS(tos), rt2->dst.dev);
525
526 dst_release(&rt2->dst);
527 rt2 = skb_rtable(skb_in);
528 skb_in->_skb_refdst = orefdst; /* restore old refdst */
529 }
530
531 if (err)
532 goto relookup_failed;
533
David S. Miller9d6ec932011-03-12 01:12:47 -0500534 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
David S. Miller415b3332011-07-22 06:22:10 -0700535 flowi4_to_flowi(&fl4_dec), NULL,
David S. Miller9d6ec932011-03-12 01:12:47 -0500536 XFRM_LOOKUP_ICMP);
David S. Miller452edd52011-03-02 13:27:41 -0800537 if (!IS_ERR(rt2)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800538 dst_release(&rt->dst);
David S. Miller415b3332011-07-22 06:22:10 -0700539 memcpy(fl4, &fl4_dec, sizeof(*fl4));
David S. Millerf6d460c2011-03-01 15:49:55 -0800540 rt = rt2;
David S. Miller452edd52011-03-02 13:27:41 -0800541 } else if (PTR_ERR(rt2) == -EPERM) {
542 if (rt)
543 dst_release(&rt->dst);
544 return rt2;
545 } else {
546 err = PTR_ERR(rt2);
547 goto relookup_failed;
David S. Millerf6d460c2011-03-01 15:49:55 -0800548 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800549 return rt;
550
551relookup_failed:
552 if (rt)
553 return rt;
554 return ERR_PTR(err);
555}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
557/*
558 * Send an ICMP message in response to a situation
559 *
560 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
561 * MAY send more (we do).
562 * MUST NOT change this header information.
563 * MUST NOT reply to a multicast/broadcast IP address.
564 * MUST NOT reply to a multicast/broadcast MAC address.
565 * MUST reply to only the first fragment.
566 */
567
Al Viroe4883012006-09-26 22:17:28 -0700568void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569{
570 struct iphdr *iph;
571 int room;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000572 struct icmp_bxm *icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000573 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 struct ipcm_cookie ipc;
David S. Miller77968b72011-05-08 17:12:19 -0700575 struct flowi4 fl4;
Al Viroa61ced52006-09-26 21:27:54 -0700576 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 u8 tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700578 u32 mark;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800579 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800580 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
582 if (!rt)
583 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700584 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
586 /*
587 * Find the original header. It is expected to be valid, of course.
588 * Check this, icmp_send is called from the most obscure devices
589 * sometimes.
590 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700591 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700593 if ((u8 *)iph < skb_in->head ||
Simon Hormanf7c0c2a2013-05-28 20:34:27 +0000594 (skb_network_header(skb_in) + sizeof(*iph)) >
595 skb_tail_pointer(skb_in))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 goto out;
597
598 /*
599 * No replies to physical multicast/broadcast
600 */
601 if (skb_in->pkt_type != PACKET_HOST)
602 goto out;
603
604 /*
605 * Now check at the protocol level
606 */
607 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
608 goto out;
609
610 /*
611 * Only reply to fragment 0. We byte re-order the constant
612 * mask for efficiency.
613 */
614 if (iph->frag_off & htons(IP_OFFSET))
615 goto out;
616
617 /*
618 * If we send an ICMP error to an ICMP error a mess would result..
619 */
620 if (icmp_pointers[type].error) {
621 /*
622 * We are an error, check if we are replying to an
623 * ICMP error
624 */
625 if (iph->protocol == IPPROTO_ICMP) {
626 u8 _inner_type, *itp;
627
628 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700629 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 (iph->ihl << 2) +
631 offsetof(struct icmphdr,
632 type) -
633 skb_in->data,
634 sizeof(_inner_type),
635 &_inner_type);
Ian Morris51456b22015-04-03 09:17:26 +0100636 if (!itp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 goto out;
638
639 /*
640 * Assume any unknown ICMP type is an error. This
641 * isn't specified by the RFC, but think about it..
642 */
643 if (*itp > NR_ICMP_TYPES ||
644 icmp_pointers[*itp].error)
645 goto out;
646 }
647 }
648
Cong Wang9a99d4a2013-06-02 15:00:52 +0000649 icmp_param = kmalloc(sizeof(*icmp_param), GFP_ATOMIC);
650 if (!icmp_param)
651 return;
652
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700653 sk = icmp_xmit_lock(net);
Ian Morris51456b22015-04-03 09:17:26 +0100654 if (!sk)
Cong Wang9a99d4a2013-06-02 15:00:52 +0000655 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
657 /*
658 * Construct source address and options.
659 */
660
661 saddr = iph->daddr;
Paolo Abeni02a6d612015-10-14 14:25:53 +0200662 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700663 struct net_device *dev = NULL;
664
Eric Dumazet685c7942009-11-01 19:31:03 +0000665 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800666 if (rt_is_input_route(rt) &&
667 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller92101b32012-07-23 16:29:00 -0700668 dev = dev_get_by_index_rcu(net, inet_iif(skb_in));
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700669
Eric Dumazet685c7942009-11-01 19:31:03 +0000670 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700671 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000672 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700673 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000674 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700675 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
677 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
678 IPTOS_PREC_INTERNETCONTROL) :
679 iph->tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700680 mark = IP4_REPLY_MARK(net, skb_in->mark);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681
Cong Wang9a99d4a2013-06-02 15:00:52 +0000682 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800683 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
685
686 /*
687 * Prepare data for ICMP header.
688 */
689
Cong Wang9a99d4a2013-06-02 15:00:52 +0000690 icmp_param->data.icmph.type = type;
691 icmp_param->data.icmph.code = code;
692 icmp_param->data.icmph.un.gateway = info;
693 icmp_param->data.icmph.checksum = 0;
694 icmp_param->skb = skb_in;
695 icmp_param->offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800696 inet_sk(sk)->tos = tos;
Lorenzo Colittie1108612014-05-13 10:17:33 -0700697 sk->sk_mark = mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 ipc.addr = iph->saddr;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000699 ipc.opt = &icmp_param->replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000700 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200701 ipc.ttl = 0;
702 ipc.tos = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703
Lorenzo Colittie1108612014-05-13 10:17:33 -0700704 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, mark,
Cong Wang9a99d4a2013-06-02 15:00:52 +0000705 type, code, icmp_param);
David S. Millerf6d460c2011-03-01 15:49:55 -0800706 if (IS_ERR(rt))
707 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800708
David S. Millera48eff12011-05-18 18:42:43 -0400709 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 goto ende;
711
712 /* RFC says return as much as we can without exceeding 576 bytes. */
713
Changli Gaod8d1f302010-06-10 23:31:35 -0700714 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 if (room > 576)
716 room = 576;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000717 room -= sizeof(struct iphdr) + icmp_param->replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 room -= sizeof(struct icmphdr);
719
Cong Wang9a99d4a2013-06-02 15:00:52 +0000720 icmp_param->data_len = skb_in->len - icmp_param->offset;
721 if (icmp_param->data_len > room)
722 icmp_param->data_len = room;
723 icmp_param->head_len = sizeof(struct icmphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724
Cong Wang9a99d4a2013-06-02 15:00:52 +0000725 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726ende:
727 ip_rt_put(rt);
728out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800729 icmp_xmit_unlock(sk);
Cong Wang9a99d4a2013-06-02 15:00:52 +0000730out_free:
731 kfree(icmp_param);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732out:;
733}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000734EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735
736
David S. Miller1de92432012-07-11 18:32:17 -0700737static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
738{
739 const struct iphdr *iph = (const struct iphdr *) skb->data;
740 const struct net_protocol *ipprot;
741 int protocol = iph->protocol;
742
David S. Millerf0a70e92012-07-12 08:06:04 -0700743 /* Checkin full IP header plus 8 bytes of protocol to
744 * avoid additional coding at protocol handlers.
745 */
Duan Jiong7304fe42014-07-31 17:54:32 +0800746 if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) {
747 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
David S. Millerf0a70e92012-07-12 08:06:04 -0700748 return;
Duan Jiong7304fe42014-07-31 17:54:32 +0800749 }
David S. Millerf0a70e92012-07-12 08:06:04 -0700750
David S. Miller1de92432012-07-11 18:32:17 -0700751 raw_icmp_error(skb, protocol, info);
752
David S. Miller1de92432012-07-11 18:32:17 -0700753 ipprot = rcu_dereference(inet_protos[protocol]);
754 if (ipprot && ipprot->err_handler)
755 ipprot->err_handler(skb, info);
David S. Miller1de92432012-07-11 18:32:17 -0700756}
757
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +0100758static bool icmp_tag_validation(int proto)
759{
760 bool ok;
761
762 rcu_read_lock();
763 ok = rcu_dereference(inet_protos[proto])->icmp_strict_tag_validation;
764 rcu_read_unlock();
765 return ok;
766}
767
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768/*
Rami Rosen08578d82013-06-03 00:23:25 +0000769 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, ICMP_QUENCH, and
770 * ICMP_PARAMETERPROB.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 */
772
Rick Jonese3e32172014-11-17 14:04:29 -0800773static bool icmp_unreach(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000775 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 struct icmphdr *icmph;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800777 struct net *net;
David S. Millerf9242b62012-06-19 18:56:21 -0700778 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800779
Eric Dumazetadf30902009-06-02 05:19:30 +0000780 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782 /*
783 * Incomplete header ?
784 * Only checks for the IP header, there should be an
785 * additional check for longer headers in upper levels.
786 */
787
788 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
789 goto out_err;
790
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300791 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000792 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794 if (iph->ihl < 5) /* Mangled header, drop. */
795 goto out_err;
796
797 if (icmph->type == ICMP_DEST_UNREACH) {
798 switch (icmph->code & 15) {
799 case ICMP_NET_UNREACH:
800 case ICMP_HOST_UNREACH:
801 case ICMP_PROT_UNREACH:
802 case ICMP_PORT_UNREACH:
803 break;
804 case ICMP_FRAG_NEEDED:
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +0100805 /* for documentation of the ip_no_pmtu_disc
806 * values please see
807 * Documentation/networking/ip-sysctl.txt
808 */
809 switch (net->ipv4.sysctl_ip_no_pmtu_disc) {
810 default:
Joe Perchesba7a46f2014-11-11 10:59:17 -0800811 net_dbg_ratelimited("%pI4: fragmentation needed and DF set\n",
812 &iph->daddr);
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +0100813 break;
814 case 2:
815 goto out;
816 case 3:
817 if (!icmp_tag_validation(iph->protocol))
818 goto out;
819 /* fall through */
820 case 0:
David S. Miller46517002012-06-10 00:04:12 -0700821 info = ntohs(icmph->un.frag.mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 }
823 break;
824 case ICMP_SR_FAILED:
Joe Perchesba7a46f2014-11-11 10:59:17 -0800825 net_dbg_ratelimited("%pI4: Source Route Failed\n",
826 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 break;
828 default:
829 break;
830 }
831 if (icmph->code > NR_ICMP_UNREACH)
832 goto out;
833 } else if (icmph->type == ICMP_PARAMETERPROB)
834 info = ntohl(icmph->un.gateway) >> 24;
835
836 /*
837 * Throw it at our lower layers
838 *
839 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
840 * header.
841 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
842 * transport layer.
843 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
844 * transport layer.
845 */
846
847 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300848 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 * bogus responses to broadcast frames. If you see this message
850 * first check your netmask matches at both ends, if it does then
851 * get the other vendor to fix their kit.
852 */
853
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700854 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
David Ahern30bbaa12015-08-13 14:59:05 -0600855 inet_addr_type_dev_table(net, skb->dev, iph->daddr) == RTN_BROADCAST) {
Joe Perchese87cc472012-05-13 21:56:26 +0000856 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
857 &ip_hdr(skb)->saddr,
858 icmph->type, icmph->code,
859 &iph->daddr, skb->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 goto out;
861 }
862
David S. Miller1de92432012-07-11 18:32:17 -0700863 icmp_socket_deliver(skb, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
865out:
Rick Jonese3e32172014-11-17 14:04:29 -0800866 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700868 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Rick Jonese3e32172014-11-17 14:04:29 -0800869 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870}
871
872
873/*
874 * Handle ICMP_REDIRECT.
875 */
876
Rick Jonese3e32172014-11-17 14:04:29 -0800877static bool icmp_redirect(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878{
David S. Miller94206122012-07-11 20:38:08 -0700879 if (skb->len < sizeof(struct iphdr)) {
880 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
Rick Jonese3e32172014-11-17 14:04:29 -0800881 return false;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900882 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000883
Rick Jonese3e32172014-11-17 14:04:29 -0800884 if (!pskb_may_pull(skb, sizeof(struct iphdr))) {
885 /* there aught to be a stat */
886 return false;
887 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000888
David S. Miller94206122012-07-11 20:38:08 -0700889 icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway);
Rick Jonese3e32172014-11-17 14:04:29 -0800890 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891}
892
893/*
894 * Handle ICMP_ECHO ("ping") requests.
895 *
896 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
897 * requests.
898 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
899 * included in the reply.
900 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
901 * echo requests, MUST have default=NOT.
902 * See also WRT handling of options once they are done and working.
903 */
904
Rick Jonese3e32172014-11-17 14:04:29 -0800905static bool icmp_echo(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700907 struct net *net;
908
Eric Dumazetadf30902009-06-02 05:19:30 +0000909 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700910 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 struct icmp_bxm icmp_param;
912
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300913 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
915 icmp_param.skb = skb;
916 icmp_param.offset = 0;
917 icmp_param.data_len = skb->len;
918 icmp_param.head_len = sizeof(struct icmphdr);
919 icmp_reply(&icmp_param, skb);
920 }
Rick Jonese3e32172014-11-17 14:04:29 -0800921 /* should there be an ICMP stat for ignored echos? */
922 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923}
924
925/*
926 * Handle ICMP Timestamp requests.
927 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
928 * SHOULD be in the kernel for minimum random latency.
929 * MUST be accurate to a few minutes.
930 * MUST be updated at least at 15Hz.
931 */
Rick Jonese3e32172014-11-17 14:04:29 -0800932static bool icmp_timestamp(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700934 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 struct icmp_bxm icmp_param;
936 /*
937 * Too short.
938 */
939 if (skb->len < 4)
940 goto out_err;
941
942 /*
943 * Fill in the current time as ms since midnight UT:
944 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700945 getnstimeofday(&tv);
946 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
947 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 icmp_param.data.times[2] = icmp_param.data.times[1];
949 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
950 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300951 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
953 icmp_param.data.icmph.code = 0;
954 icmp_param.skb = skb;
955 icmp_param.offset = 0;
956 icmp_param.data_len = 0;
957 icmp_param.head_len = sizeof(struct icmphdr) + 12;
958 icmp_reply(&icmp_param, skb);
Rick Jonese3e32172014-11-17 14:04:29 -0800959 return true;
960
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000962 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Rick Jonese3e32172014-11-17 14:04:29 -0800963 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964}
965
Rick Jonese3e32172014-11-17 14:04:29 -0800966static bool icmp_discard(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967{
Rick Jonese3e32172014-11-17 14:04:29 -0800968 /* pretend it was a success */
969 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970}
971
972/*
973 * Deal with incoming ICMP packets.
974 */
975int icmp_rcv(struct sk_buff *skb)
976{
977 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000978 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700979 struct net *net = dev_net(rt->dst.dev);
Rick Jonese3e32172014-11-17 14:04:29 -0800980 bool success;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981
Herbert Xuaebcf822007-12-12 18:54:16 -0800982 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700983 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800984 int nh;
985
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700986 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800987 XFRM_STATE_ICMP))
988 goto drop;
989
Herbert Xu8b7817f2007-12-12 10:44:43 -0800990 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
991 goto drop;
992
993 nh = skb_network_offset(skb);
994 skb_set_network_header(skb, sizeof(*icmph));
995
996 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
997 goto drop;
998
999 skb_set_network_header(skb, nh);
1000 }
1001
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001002 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Tom Herbert29a96e12014-05-07 16:52:21 -07001004 if (skb_checksum_simple_validate(skb))
1005 goto csum_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Herbert Xu8cf22942008-02-05 03:15:50 -08001007 if (!pskb_pull(skb, sizeof(*icmph)))
1008 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -03001010 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
Pavel Emelyanovf66ac032008-07-14 23:04:00 -07001012 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 /*
1014 * 18 is the highest 'known' ICMP type. Anything else is a mystery
1015 *
1016 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
1017 * discarded.
1018 */
1019 if (icmph->type > NR_ICMP_TYPES)
1020 goto error;
1021
1022
1023 /*
1024 * Parse the ICMP message
1025 */
1026
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001027 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 /*
1029 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
1030 * silently ignored (we let user decide with a sysctl).
1031 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1032 * discarded if to broadcast/multicast.
1033 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -07001034 if ((icmph->type == ICMP_ECHO ||
1035 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -07001036 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 goto error;
1038 }
1039 if (icmph->type != ICMP_ECHO &&
1040 icmph->type != ICMP_TIMESTAMP &&
1041 icmph->type != ICMP_ADDRESS &&
1042 icmph->type != ICMP_ADDRESSREPLY) {
1043 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001044 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 }
1046
Rick Jonese3e32172014-11-17 14:04:29 -08001047 success = icmp_pointers[icmph->type].handler(skb);
1048
1049 if (success) {
1050 consume_skb(skb);
1051 return 0;
1052 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053
1054drop:
1055 kfree_skb(skb);
1056 return 0;
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001057csum_error:
1058 ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -07001060 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 goto drop;
1062}
1063
Li Wei5b052042013-02-21 22:18:44 +00001064void icmp_err(struct sk_buff *skb, u32 info)
1065{
1066 struct iphdr *iph = (struct iphdr *)skb->data;
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +00001067 int offset = iph->ihl<<2;
1068 struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset);
Li Wei5b052042013-02-21 22:18:44 +00001069 int type = icmp_hdr(skb)->type;
1070 int code = icmp_hdr(skb)->code;
1071 struct net *net = dev_net(skb->dev);
1072
1073 /*
1074 * Use ping_err to handle all icmp errors except those
1075 * triggered by ICMP_ECHOREPLY which sent from kernel.
1076 */
1077 if (icmph->type != ICMP_ECHOREPLY) {
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +00001078 ping_err(skb, offset, info);
Li Wei5b052042013-02-21 22:18:44 +00001079 return;
1080 }
1081
1082 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
1083 ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0);
1084 else if (type == ICMP_REDIRECT)
1085 ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0);
1086}
1087
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088/*
1089 * This table is the definition of how we handle ICMP.
1090 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -08001091static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001093 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 },
1095 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 .handler = icmp_discard,
1097 .error = 1,
1098 },
1099 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 .handler = icmp_discard,
1101 .error = 1,
1102 },
1103 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 .handler = icmp_unreach,
1105 .error = 1,
1106 },
1107 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 .handler = icmp_unreach,
1109 .error = 1,
1110 },
1111 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 .handler = icmp_redirect,
1113 .error = 1,
1114 },
1115 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 .handler = icmp_discard,
1117 .error = 1,
1118 },
1119 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 .handler = icmp_discard,
1121 .error = 1,
1122 },
1123 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 .handler = icmp_echo,
1125 },
1126 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 .handler = icmp_discard,
1128 .error = 1,
1129 },
1130 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 .handler = icmp_discard,
1132 .error = 1,
1133 },
1134 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 .handler = icmp_unreach,
1136 .error = 1,
1137 },
1138 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 .handler = icmp_unreach,
1140 .error = 1,
1141 },
1142 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 .handler = icmp_timestamp,
1144 },
1145 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 .handler = icmp_discard,
1147 },
1148 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 .handler = icmp_discard,
1150 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001151 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 .handler = icmp_discard,
1153 },
1154 [ICMP_ADDRESS] = {
David S. Miller838942a2012-07-23 13:20:26 -07001155 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156 },
1157 [ICMP_ADDRESSREPLY] = {
David S. Miller838942a2012-07-23 13:20:26 -07001158 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 },
1160};
1161
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001162static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 int i;
1165
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001166 for_each_possible_cpu(i)
Eric Dumazet349c9e32015-01-29 15:58:09 -08001167 inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i));
1168 free_percpu(net->ipv4.icmp_sk);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001169 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001170}
1171
Adrian Bunk263173a2008-04-21 02:31:23 -07001172static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001173{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001174 int i, err;
1175
Eric Dumazet349c9e32015-01-29 15:58:09 -08001176 net->ipv4.icmp_sk = alloc_percpu(struct sock *);
1177 if (!net->ipv4.icmp_sk)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001178 return -ENOMEM;
1179
Denis V. Luneva5710d62008-02-29 11:14:50 -08001180 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001181 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001183 err = inet_ctl_sock_create(&sk, PF_INET,
1184 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001186 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187
Eric Dumazet349c9e32015-01-29 15:58:09 -08001188 *per_cpu_ptr(net->ipv4.icmp_sk, i) = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189
1190 /* Enough space for 2 64K ICMP packets, including
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001191 * sk_buff/skb_shared_info struct overhead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 */
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001193 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001195 /*
1196 * Speedup sock_wfree()
1197 */
1198 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001199 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001201
1202 /* Control parameters for ECHO replies. */
1203 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1204 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1205
1206 /* Control parameter - ignore bogus broadcast responses? */
1207 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1208
1209 /*
1210 * Configurable global rate limit.
1211 *
1212 * ratelimit defines tokens/packet consumed for dst->rate_token
1213 * bucket ratemask defines which icmp types are ratelimited by
1214 * setting it's bit position.
1215 *
1216 * default:
1217 * dest unreachable (3), source quench (4),
1218 * time exceeded (11), parameter problem (12)
1219 */
1220
1221 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1222 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1223 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1224
Denis V. Luneva5710d62008-02-29 11:14:50 -08001225 return 0;
1226
1227fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001228 for_each_possible_cpu(i)
Eric Dumazet349c9e32015-01-29 15:58:09 -08001229 inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i));
1230 free_percpu(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001231 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232}
1233
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001234static struct pernet_operations __net_initdata icmp_sk_ops = {
1235 .init = icmp_sk_init,
1236 .exit = icmp_sk_exit,
1237};
1238
1239int __init icmp_init(void)
1240{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001241 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001242}