blob: 5d0d379b0152d76df92f2cfaaf60799358d5191e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET3: Implementation of the ICMP protocol layer.
3 *
Alan Cox113aa832008-10-13 19:01:08 -07004 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07006 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 * Some of the function names and the icmp unreach table for this
12 * module were derived from [icmp.c 1.0.11 06/02/93] by
13 * Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
14 * Other than that this module is a complete rewrite.
15 *
16 * Fixes:
17 * Clemens Fruhwirth : introduce global icmp rate limiting
18 * with icmp type masking ability instead
19 * of broken per type icmp timeouts.
20 * Mike Shaver : RFC1122 checks.
21 * Alan Cox : Multicast ping reply as self.
22 * Alan Cox : Fix atomicity lockup in ip_build_xmit
23 * call.
24 * Alan Cox : Added 216,128 byte paths to the MTU
25 * code.
26 * Martin Mares : RFC1812 checks.
27 * Martin Mares : Can be configured to follow redirects
28 * if acting as a router _without_ a
29 * routing protocol (RFC 1812).
30 * Martin Mares : Echo requests may be configured to
31 * be ignored (RFC 1812).
32 * Martin Mares : Limitation of ICMP error message
33 * transmit rate (RFC 1812).
34 * Martin Mares : TOS and Precedence set correctly
35 * (RFC 1812).
36 * Martin Mares : Now copying as much data from the
37 * original packet as we can without
38 * exceeding 576 bytes (RFC 1812).
39 * Willy Konynenberg : Transparent proxying support.
40 * Keith Owens : RFC1191 correction for 4.2BSD based
41 * path MTU bug.
42 * Thomas Quinot : ICMP Dest Unreach codes up to 15 are
43 * valid (RFC 1812).
44 * Andi Kleen : Check all packet lengths properly
45 * and moved all kfree_skb() up to
46 * icmp_rcv.
47 * Andi Kleen : Move the rate limit bookkeeping
48 * into the dest entry and use a token
49 * bucket filter (thanks to ANK). Make
50 * the rates sysctl configurable.
51 * Yu Tianli : Fixed two ugly bugs in icmp_send
52 * - IP option length was accounted wrongly
53 * - ICMP header length was not accounted
54 * at all.
55 * Tristan Greaves : Added sysctl option to ignore bogus
56 * broadcast responses from broken routers.
57 *
58 * To Fix:
59 *
60 * - Should use skb_pull() instead of all the manual checking.
61 * This would also greatly simply some upper layer error handlers. --AK
62 *
63 */
64
Joe Perchesafd465032012-03-12 07:03:32 +000065#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/module.h>
68#include <linux/types.h>
69#include <linux/jiffies.h>
70#include <linux/kernel.h>
71#include <linux/fcntl.h>
72#include <linux/socket.h>
73#include <linux/in.h>
74#include <linux/inet.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020075#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/netdevice.h>
77#include <linux/string.h>
78#include <linux/netfilter_ipv4.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090079#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#include <net/snmp.h>
81#include <net/ip.h>
82#include <net/route.h>
83#include <net/protocol.h>
84#include <net/icmp.h>
85#include <net/tcp.h>
86#include <net/udp.h>
87#include <net/raw.h>
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +000088#include <net/ping.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070089#include <linux/skbuff.h>
90#include <net/sock.h>
91#include <linux/errno.h>
92#include <linux/timer.h>
93#include <linux/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include <asm/uaccess.h>
95#include <net/checksum.h>
Herbert Xu8b7817f2007-12-12 10:44:43 -080096#include <net/xfrm.h>
Denis V. Lunevc1e98942008-04-03 14:29:00 -070097#include <net/inet_common.h>
David S. Miller35ebf652012-06-28 03:59:11 -070098#include <net/ip_fib.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100/*
101 * Build xmit assembly blocks
102 */
103
104struct icmp_bxm {
105 struct sk_buff *skb;
106 int offset;
107 int data_len;
108
109 struct {
110 struct icmphdr icmph;
Al Virob03d73e2006-09-27 18:38:30 -0700111 __be32 times[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 } data;
113 int head_len;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000114 struct ip_options_data replyopts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
116
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117/* An array of errno for error messages from dest unreach. */
118/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
119
Alexey Dobriyane7548342010-01-22 10:18:25 +0000120const struct icmp_err icmp_err_convert[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121 {
122 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
123 .fatal = 0,
124 },
125 {
126 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
127 .fatal = 0,
128 },
129 {
130 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
131 .fatal = 1,
132 },
133 {
134 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
135 .fatal = 1,
136 },
137 {
138 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
139 .fatal = 0,
140 },
141 {
142 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
143 .fatal = 0,
144 },
145 {
146 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
147 .fatal = 1,
148 },
149 {
150 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
151 .fatal = 1,
152 },
153 {
154 .errno = ENONET, /* ICMP_HOST_ISOLATED */
155 .fatal = 1,
156 },
157 {
158 .errno = ENETUNREACH, /* ICMP_NET_ANO */
159 .fatal = 1,
160 },
161 {
162 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
163 .fatal = 1,
164 },
165 {
166 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
167 .fatal = 0,
168 },
169 {
170 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
171 .fatal = 0,
172 },
173 {
174 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
175 .fatal = 1,
176 },
177 {
178 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
179 .fatal = 1,
180 },
181 {
182 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
183 .fatal = 1,
184 },
185};
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000186EXPORT_SYMBOL(icmp_err_convert);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188/*
189 * ICMP control array. This specifies what to do with each ICMP.
190 */
191
192struct icmp_control {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 void (*handler)(struct sk_buff *skb);
194 short error; /* This ICMP is classed as an error message */
195};
196
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800197static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198
199/*
200 * The ICMP socket(s). This is the most convenient way to flow control
201 * our ICMP output as well as maintain a clean interface throughout
202 * all layers. All Socketless IP sends will soon be gone.
203 *
204 * On SMP we have one ICMP socket per-cpu.
205 */
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800206static struct sock *icmp_sk(struct net *net)
207{
208 return net->ipv4.icmp_sk[smp_processor_id()];
209}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700211static inline struct sock *icmp_xmit_lock(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212{
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700213 struct sock *sk;
214
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 local_bh_disable();
216
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700217 sk = icmp_sk(net);
218
Denis V. Lunev405666d2008-02-29 11:16:46 -0800219 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 /* This can happen if the output path signals a
221 * dst_link_failure() for an outgoing ICMP packet.
222 */
223 local_bh_enable();
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700224 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 }
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700226 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227}
228
Denis V. Lunev405666d2008-02-29 11:16:46 -0800229static inline void icmp_xmit_unlock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230{
Denis V. Lunev405666d2008-02-29 11:16:46 -0800231 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232}
233
234/*
235 * Send an ICMP frame.
236 */
237
David S. Miller92d86822011-02-04 15:55:25 -0800238static inline bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
David S. Millera48eff12011-05-18 18:42:43 -0400239 struct flowi4 *fl4, int type, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240{
Changli Gaod8d1f302010-06-10 23:31:35 -0700241 struct dst_entry *dst = &rt->dst;
David S. Miller92d86822011-02-04 15:55:25 -0800242 bool rc = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
244 if (type > NR_ICMP_TYPES)
245 goto out;
246
247 /* Don't limit PMTU discovery. */
248 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
249 goto out;
250
251 /* No rate limit on loopback */
252 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900253 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
255 /* Limit if icmp type is enabled in ratemask. */
David S. Miller92d86822011-02-04 15:55:25 -0800256 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask) {
David S. Miller1d861aa2012-07-10 03:58:16 -0700257 struct inet_peer *peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, 1);
David S. Millerfbfe95a2012-06-08 23:24:18 -0700258 rc = inet_peer_xrlim_allow(peer,
David S. Miller92d86822011-02-04 15:55:25 -0800259 net->ipv4.sysctl_icmp_ratelimit);
Neal Cardwelle1a67642012-11-24 18:54:37 +0000260 if (peer)
261 inet_putpeer(peer);
David S. Miller92d86822011-02-04 15:55:25 -0800262 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263out:
264 return rc;
265}
266
267/*
268 * Maintain the counters used in the SNMP statistics for outgoing ICMP
269 */
Pavel Emelyanov0388b002008-07-14 23:00:43 -0700270void icmp_out_count(struct net *net, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271{
Pavel Emelyanov903fc192008-07-14 23:03:35 -0700272 ICMPMSGOUT_INC_STATS(net, type);
Pavel Emelyanov75c939b2008-07-14 23:02:35 -0700273 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274}
275
276/*
277 * Checksum each fragment, and on the first include the headers and final
278 * checksum.
279 */
280static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
281 struct sk_buff *skb)
282{
283 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
Al Viro5f92a732006-11-14 21:36:54 -0800284 __wsum csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
286 csum = skb_copy_and_csum_bits(icmp_param->skb,
287 icmp_param->offset + offset,
288 to, len, 0);
289
290 skb->csum = csum_block_add(skb->csum, csum, odd);
291 if (icmp_pointers[icmp_param->data.icmph.type].error)
292 nf_ct_attach(skb, icmp_param->skb);
293 return 0;
294}
295
296static void icmp_push_reply(struct icmp_bxm *icmp_param,
David S. Miller77968b72011-05-08 17:12:19 -0700297 struct flowi4 *fl4,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800298 struct ipcm_cookie *ipc, struct rtable **rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299{
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800300 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 struct sk_buff *skb;
302
Changli Gaod8d1f302010-06-10 23:31:35 -0700303 sk = icmp_sk(dev_net((*rt)->dst.dev));
David S. Millerf5fca602011-05-08 17:24:10 -0700304 if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900305 icmp_param->data_len+icmp_param->head_len,
306 icmp_param->head_len,
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700307 ipc, rt, MSG_DONTWAIT) < 0) {
308 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800309 ip_flush_pending_frames(sk);
Eric Dumazet1f8438a2010-04-03 15:09:04 -0700310 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300311 struct icmphdr *icmph = icmp_hdr(skb);
Al Virod3bc23e2006-11-14 21:24:49 -0800312 __wsum csum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 struct sk_buff *skb1;
314
Denis V. Lunev1e3cf682008-02-29 11:15:42 -0800315 skb_queue_walk(&sk->sk_write_queue, skb1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 csum = csum_add(csum, skb1->csum);
317 }
318 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
319 (char *)icmph,
320 icmp_param->head_len, csum);
321 icmph->checksum = csum_fold(csum);
322 skb->ip_summed = CHECKSUM_NONE;
David S. Miller77968b72011-05-08 17:12:19 -0700323 ip_push_pending_frames(sk, fl4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 }
325}
326
327/*
328 * Driving logic for building and sending ICMP messages.
329 */
330
331static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
332{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 struct ipcm_cookie ipc;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000334 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700335 struct net *net = dev_net(rt->dst.dev);
David S. Miller77968b72011-05-08 17:12:19 -0700336 struct flowi4 fl4;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700337 struct sock *sk;
338 struct inet_sock *inet;
David S. Miller35ebf652012-06-28 03:59:11 -0700339 __be32 daddr, saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000341 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb))
Hormsf00c4012006-02-02 17:03:18 -0800342 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700344 sk = icmp_xmit_lock(net);
345 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 return;
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700347 inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
349 icmp_param->data.icmph.checksum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700351 inet->tos = ip_hdr(skb)->tos;
David S. Miller9f6abb52011-05-09 13:28:22 -0700352 daddr = ipc.addr = ip_hdr(skb)->saddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700353 saddr = fib_compute_spec_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000355 ipc.tx_flags = 0;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000356 if (icmp_param->replyopts.opt.opt.optlen) {
357 ipc.opt = &icmp_param->replyopts.opt;
358 if (ipc.opt->opt.srr)
359 daddr = icmp_param->replyopts.opt.opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 }
David S. Miller77968b72011-05-08 17:12:19 -0700361 memset(&fl4, 0, sizeof(fl4));
362 fl4.daddr = daddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700363 fl4.saddr = saddr;
David S. Miller77968b72011-05-08 17:12:19 -0700364 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
365 fl4.flowi4_proto = IPPROTO_ICMP;
366 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
367 rt = ip_route_output_key(net, &fl4);
368 if (IS_ERR(rt))
369 goto out_unlock;
David S. Millera48eff12011-05-18 18:42:43 -0400370 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 icmp_param->data.icmph.code))
David S. Miller77968b72011-05-08 17:12:19 -0700372 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 ip_rt_put(rt);
374out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800375 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376}
377
David S. Miller77968b72011-05-08 17:12:19 -0700378static struct rtable *icmp_route_lookup(struct net *net,
379 struct flowi4 *fl4,
380 struct sk_buff *skb_in,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000381 const struct iphdr *iph,
David S. Millerf6d460c2011-03-01 15:49:55 -0800382 __be32 saddr, u8 tos,
383 int type, int code,
384 struct icmp_bxm *param)
385{
David S. Millerf6d460c2011-03-01 15:49:55 -0800386 struct rtable *rt, *rt2;
David S. Miller415b3332011-07-22 06:22:10 -0700387 struct flowi4 fl4_dec;
David S. Millerf6d460c2011-03-01 15:49:55 -0800388 int err;
389
David S. Miller77968b72011-05-08 17:12:19 -0700390 memset(fl4, 0, sizeof(*fl4));
391 fl4->daddr = (param->replyopts.opt.opt.srr ?
392 param->replyopts.opt.opt.faddr : iph->saddr);
393 fl4->saddr = saddr;
394 fl4->flowi4_tos = RT_TOS(tos);
395 fl4->flowi4_proto = IPPROTO_ICMP;
396 fl4->fl4_icmp_type = type;
397 fl4->fl4_icmp_code = code;
398 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
399 rt = __ip_route_output_key(net, fl4);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800400 if (IS_ERR(rt))
401 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800402
403 /* No need to clone since we're just using its address. */
404 rt2 = rt;
405
David S. Miller9d6ec932011-03-12 01:12:47 -0500406 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
David S. Miller77968b72011-05-08 17:12:19 -0700407 flowi4_to_flowi(fl4), NULL, 0);
David S. Miller452edd52011-03-02 13:27:41 -0800408 if (!IS_ERR(rt)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800409 if (rt != rt2)
410 return rt;
David S. Miller452edd52011-03-02 13:27:41 -0800411 } else if (PTR_ERR(rt) == -EPERM) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800412 rt = NULL;
David S. Miller452edd52011-03-02 13:27:41 -0800413 } else
414 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800415
David S. Miller415b3332011-07-22 06:22:10 -0700416 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
David S. Millerf6d460c2011-03-01 15:49:55 -0800417 if (err)
418 goto relookup_failed;
419
David S. Miller415b3332011-07-22 06:22:10 -0700420 if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) {
421 rt2 = __ip_route_output_key(net, &fl4_dec);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800422 if (IS_ERR(rt2))
423 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800424 } else {
David S. Miller9d6ec932011-03-12 01:12:47 -0500425 struct flowi4 fl4_2 = {};
David S. Millerf6d460c2011-03-01 15:49:55 -0800426 unsigned long orefdst;
427
David S. Miller415b3332011-07-22 06:22:10 -0700428 fl4_2.daddr = fl4_dec.saddr;
David S. Miller9d6ec932011-03-12 01:12:47 -0500429 rt2 = ip_route_output_key(net, &fl4_2);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800430 if (IS_ERR(rt2)) {
431 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800432 goto relookup_failed;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800433 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800434 /* Ugh! */
435 orefdst = skb_in->_skb_refdst; /* save old refdst */
David S. Miller415b3332011-07-22 06:22:10 -0700436 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
David S. Millerf6d460c2011-03-01 15:49:55 -0800437 RT_TOS(tos), rt2->dst.dev);
438
439 dst_release(&rt2->dst);
440 rt2 = skb_rtable(skb_in);
441 skb_in->_skb_refdst = orefdst; /* restore old refdst */
442 }
443
444 if (err)
445 goto relookup_failed;
446
David S. Miller9d6ec932011-03-12 01:12:47 -0500447 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
David S. Miller415b3332011-07-22 06:22:10 -0700448 flowi4_to_flowi(&fl4_dec), NULL,
David S. Miller9d6ec932011-03-12 01:12:47 -0500449 XFRM_LOOKUP_ICMP);
David S. Miller452edd52011-03-02 13:27:41 -0800450 if (!IS_ERR(rt2)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800451 dst_release(&rt->dst);
David S. Miller415b3332011-07-22 06:22:10 -0700452 memcpy(fl4, &fl4_dec, sizeof(*fl4));
David S. Millerf6d460c2011-03-01 15:49:55 -0800453 rt = rt2;
David S. Miller452edd52011-03-02 13:27:41 -0800454 } else if (PTR_ERR(rt2) == -EPERM) {
455 if (rt)
456 dst_release(&rt->dst);
457 return rt2;
458 } else {
459 err = PTR_ERR(rt2);
460 goto relookup_failed;
David S. Millerf6d460c2011-03-01 15:49:55 -0800461 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800462 return rt;
463
464relookup_failed:
465 if (rt)
466 return rt;
467 return ERR_PTR(err);
468}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470/*
471 * Send an ICMP message in response to a situation
472 *
473 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
474 * MAY send more (we do).
475 * MUST NOT change this header information.
476 * MUST NOT reply to a multicast/broadcast IP address.
477 * MUST NOT reply to a multicast/broadcast MAC address.
478 * MUST reply to only the first fragment.
479 */
480
Al Viroe4883012006-09-26 22:17:28 -0700481void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482{
483 struct iphdr *iph;
484 int room;
485 struct icmp_bxm icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000486 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 struct ipcm_cookie ipc;
David S. Miller77968b72011-05-08 17:12:19 -0700488 struct flowi4 fl4;
Al Viroa61ced52006-09-26 21:27:54 -0700489 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 u8 tos;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800491 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800492 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
494 if (!rt)
495 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700496 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497
498 /*
499 * Find the original header. It is expected to be valid, of course.
500 * Check this, icmp_send is called from the most obscure devices
501 * sometimes.
502 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700503 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700505 if ((u8 *)iph < skb_in->head ||
Simon Hormanf7c0c2a2013-05-28 20:34:27 +0000506 (skb_network_header(skb_in) + sizeof(*iph)) >
507 skb_tail_pointer(skb_in))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 goto out;
509
510 /*
511 * No replies to physical multicast/broadcast
512 */
513 if (skb_in->pkt_type != PACKET_HOST)
514 goto out;
515
516 /*
517 * Now check at the protocol level
518 */
519 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
520 goto out;
521
522 /*
523 * Only reply to fragment 0. We byte re-order the constant
524 * mask for efficiency.
525 */
526 if (iph->frag_off & htons(IP_OFFSET))
527 goto out;
528
529 /*
530 * If we send an ICMP error to an ICMP error a mess would result..
531 */
532 if (icmp_pointers[type].error) {
533 /*
534 * We are an error, check if we are replying to an
535 * ICMP error
536 */
537 if (iph->protocol == IPPROTO_ICMP) {
538 u8 _inner_type, *itp;
539
540 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700541 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 (iph->ihl << 2) +
543 offsetof(struct icmphdr,
544 type) -
545 skb_in->data,
546 sizeof(_inner_type),
547 &_inner_type);
548 if (itp == NULL)
549 goto out;
550
551 /*
552 * Assume any unknown ICMP type is an error. This
553 * isn't specified by the RFC, but think about it..
554 */
555 if (*itp > NR_ICMP_TYPES ||
556 icmp_pointers[*itp].error)
557 goto out;
558 }
559 }
560
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700561 sk = icmp_xmit_lock(net);
562 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 return;
564
565 /*
566 * Construct source address and options.
567 */
568
569 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700570 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700571 struct net_device *dev = NULL;
572
Eric Dumazet685c7942009-11-01 19:31:03 +0000573 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800574 if (rt_is_input_route(rt) &&
575 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller92101b32012-07-23 16:29:00 -0700576 dev = dev_get_by_index_rcu(net, inet_iif(skb_in));
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700577
Eric Dumazet685c7942009-11-01 19:31:03 +0000578 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700579 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000580 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700581 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000582 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700583 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
585 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
586 IPTOS_PREC_INTERNETCONTROL) :
587 iph->tos;
588
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000589 if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800590 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591
592
593 /*
594 * Prepare data for ICMP header.
595 */
596
597 icmp_param.data.icmph.type = type;
598 icmp_param.data.icmph.code = code;
599 icmp_param.data.icmph.un.gateway = info;
600 icmp_param.data.icmph.checksum = 0;
601 icmp_param.skb = skb_in;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700602 icmp_param.offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800603 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 ipc.addr = iph->saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000605 ipc.opt = &icmp_param.replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000606 ipc.tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
David S. Miller77968b72011-05-08 17:12:19 -0700608 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos,
David S. Millerf6d460c2011-03-01 15:49:55 -0800609 type, code, &icmp_param);
610 if (IS_ERR(rt))
611 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800612
David S. Millera48eff12011-05-18 18:42:43 -0400613 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 goto ende;
615
616 /* RFC says return as much as we can without exceeding 576 bytes. */
617
Changli Gaod8d1f302010-06-10 23:31:35 -0700618 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 if (room > 576)
620 room = 576;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000621 room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 room -= sizeof(struct icmphdr);
623
624 icmp_param.data_len = skb_in->len - icmp_param.offset;
625 if (icmp_param.data_len > room)
626 icmp_param.data_len = room;
627 icmp_param.head_len = sizeof(struct icmphdr);
628
David S. Miller77968b72011-05-08 17:12:19 -0700629 icmp_push_reply(&icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630ende:
631 ip_rt_put(rt);
632out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800633 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634out:;
635}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000636EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637
638
David S. Miller1de92432012-07-11 18:32:17 -0700639static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
640{
641 const struct iphdr *iph = (const struct iphdr *) skb->data;
642 const struct net_protocol *ipprot;
643 int protocol = iph->protocol;
644
David S. Millerf0a70e92012-07-12 08:06:04 -0700645 /* Checkin full IP header plus 8 bytes of protocol to
646 * avoid additional coding at protocol handlers.
647 */
648 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
649 return;
650
David S. Miller1de92432012-07-11 18:32:17 -0700651 raw_icmp_error(skb, protocol, info);
652
653 rcu_read_lock();
654 ipprot = rcu_dereference(inet_protos[protocol]);
655 if (ipprot && ipprot->err_handler)
656 ipprot->err_handler(skb, info);
657 rcu_read_unlock();
658}
659
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660/*
661 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
662 */
663
664static void icmp_unreach(struct sk_buff *skb)
665{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000666 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 struct icmphdr *icmph;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800668 struct net *net;
David S. Millerf9242b62012-06-19 18:56:21 -0700669 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800670
Eric Dumazetadf30902009-06-02 05:19:30 +0000671 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
673 /*
674 * Incomplete header ?
675 * Only checks for the IP header, there should be an
676 * additional check for longer headers in upper levels.
677 */
678
679 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
680 goto out_err;
681
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300682 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000683 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
685 if (iph->ihl < 5) /* Mangled header, drop. */
686 goto out_err;
687
688 if (icmph->type == ICMP_DEST_UNREACH) {
689 switch (icmph->code & 15) {
690 case ICMP_NET_UNREACH:
691 case ICMP_HOST_UNREACH:
692 case ICMP_PROT_UNREACH:
693 case ICMP_PORT_UNREACH:
694 break;
695 case ICMP_FRAG_NEEDED:
696 if (ipv4_config.no_pmtu_disc) {
Joe Perchesafd465032012-03-12 07:03:32 +0000697 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700698 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 } else {
David S. Miller46517002012-06-10 00:04:12 -0700700 info = ntohs(icmph->un.frag.mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 if (!info)
702 goto out;
703 }
704 break;
705 case ICMP_SR_FAILED:
Joe Perchesafd465032012-03-12 07:03:32 +0000706 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700707 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 break;
709 default:
710 break;
711 }
712 if (icmph->code > NR_ICMP_UNREACH)
713 goto out;
714 } else if (icmph->type == ICMP_PARAMETERPROB)
715 info = ntohl(icmph->un.gateway) >> 24;
716
717 /*
718 * Throw it at our lower layers
719 *
720 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
721 * header.
722 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
723 * transport layer.
724 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
725 * transport layer.
726 */
727
728 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300729 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 * bogus responses to broadcast frames. If you see this message
731 * first check your netmask matches at both ends, if it does then
732 * get the other vendor to fix their kit.
733 */
734
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700735 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800736 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Joe Perchese87cc472012-05-13 21:56:26 +0000737 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
738 &ip_hdr(skb)->saddr,
739 icmph->type, icmph->code,
740 &iph->daddr, skb->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 goto out;
742 }
743
David S. Miller1de92432012-07-11 18:32:17 -0700744 icmp_socket_deliver(skb, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745
746out:
747 return;
748out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700749 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 goto out;
751}
752
753
754/*
755 * Handle ICMP_REDIRECT.
756 */
757
758static void icmp_redirect(struct sk_buff *skb)
759{
David S. Miller94206122012-07-11 20:38:08 -0700760 if (skb->len < sizeof(struct iphdr)) {
761 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
762 return;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900763 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000764
David S. Miller94206122012-07-11 20:38:08 -0700765 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
766 return;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000767
David S. Miller94206122012-07-11 20:38:08 -0700768 icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769}
770
771/*
772 * Handle ICMP_ECHO ("ping") requests.
773 *
774 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
775 * requests.
776 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
777 * included in the reply.
778 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
779 * echo requests, MUST have default=NOT.
780 * See also WRT handling of options once they are done and working.
781 */
782
783static void icmp_echo(struct sk_buff *skb)
784{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700785 struct net *net;
786
Eric Dumazetadf30902009-06-02 05:19:30 +0000787 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700788 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 struct icmp_bxm icmp_param;
790
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300791 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
793 icmp_param.skb = skb;
794 icmp_param.offset = 0;
795 icmp_param.data_len = skb->len;
796 icmp_param.head_len = sizeof(struct icmphdr);
797 icmp_reply(&icmp_param, skb);
798 }
799}
800
801/*
802 * Handle ICMP Timestamp requests.
803 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
804 * SHOULD be in the kernel for minimum random latency.
805 * MUST be accurate to a few minutes.
806 * MUST be updated at least at 15Hz.
807 */
808static void icmp_timestamp(struct sk_buff *skb)
809{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700810 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 struct icmp_bxm icmp_param;
812 /*
813 * Too short.
814 */
815 if (skb->len < 4)
816 goto out_err;
817
818 /*
819 * Fill in the current time as ms since midnight UT:
820 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700821 getnstimeofday(&tv);
822 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
823 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 icmp_param.data.times[2] = icmp_param.data.times[1];
825 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
826 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300827 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
829 icmp_param.data.icmph.code = 0;
830 icmp_param.skb = skb;
831 icmp_param.offset = 0;
832 icmp_param.data_len = 0;
833 icmp_param.head_len = sizeof(struct icmphdr) + 12;
834 icmp_reply(&icmp_param, skb);
835out:
836 return;
837out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000838 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 goto out;
840}
841
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842static void icmp_discard(struct sk_buff *skb)
843{
844}
845
846/*
847 * Deal with incoming ICMP packets.
848 */
849int icmp_rcv(struct sk_buff *skb)
850{
851 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000852 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700853 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854
Herbert Xuaebcf822007-12-12 18:54:16 -0800855 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700856 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800857 int nh;
858
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700859 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800860 XFRM_STATE_ICMP))
861 goto drop;
862
Herbert Xu8b7817f2007-12-12 10:44:43 -0800863 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
864 goto drop;
865
866 nh = skb_network_offset(skb);
867 skb_set_network_header(skb, sizeof(*icmph));
868
869 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
870 goto drop;
871
872 skb_set_network_header(skb, nh);
873 }
874
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700875 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876
877 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700878 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800879 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 break;
Herbert Xufb286bb2005-11-10 13:01:24 -0800881 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -0800883 skb->csum = 0;
884 if (__skb_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000885 goto csum_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 }
887
Herbert Xu8cf22942008-02-05 03:15:50 -0800888 if (!pskb_pull(skb, sizeof(*icmph)))
889 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300891 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Pavel Emelyanovf66ac032008-07-14 23:04:00 -0700893 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 /*
895 * 18 is the highest 'known' ICMP type. Anything else is a mystery
896 *
897 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
898 * discarded.
899 */
900 if (icmph->type > NR_ICMP_TYPES)
901 goto error;
902
903
904 /*
905 * Parse the ICMP message
906 */
907
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900908 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 /*
910 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
911 * silently ignored (we let user decide with a sysctl).
912 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
913 * discarded if to broadcast/multicast.
914 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -0700915 if ((icmph->type == ICMP_ECHO ||
916 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700917 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 goto error;
919 }
920 if (icmph->type != ICMP_ECHO &&
921 icmph->type != ICMP_TIMESTAMP &&
922 icmph->type != ICMP_ADDRESS &&
923 icmph->type != ICMP_ADDRESSREPLY) {
924 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900925 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 }
927
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 icmp_pointers[icmph->type].handler(skb);
929
930drop:
931 kfree_skb(skb);
932 return 0;
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000933csum_error:
934 ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700936 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 goto drop;
938}
939
Li Wei5b052042013-02-21 22:18:44 +0000940void icmp_err(struct sk_buff *skb, u32 info)
941{
942 struct iphdr *iph = (struct iphdr *)skb->data;
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +0000943 int offset = iph->ihl<<2;
944 struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset);
Li Wei5b052042013-02-21 22:18:44 +0000945 int type = icmp_hdr(skb)->type;
946 int code = icmp_hdr(skb)->code;
947 struct net *net = dev_net(skb->dev);
948
949 /*
950 * Use ping_err to handle all icmp errors except those
951 * triggered by ICMP_ECHOREPLY which sent from kernel.
952 */
953 if (icmph->type != ICMP_ECHOREPLY) {
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +0000954 ping_err(skb, offset, info);
Li Wei5b052042013-02-21 22:18:44 +0000955 return;
956 }
957
958 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
959 ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0);
960 else if (type == ICMP_REDIRECT)
961 ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0);
962}
963
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964/*
965 * This table is the definition of how we handle ICMP.
966 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800967static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000969 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 },
971 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 .handler = icmp_discard,
973 .error = 1,
974 },
975 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 .handler = icmp_discard,
977 .error = 1,
978 },
979 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 .handler = icmp_unreach,
981 .error = 1,
982 },
983 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 .handler = icmp_unreach,
985 .error = 1,
986 },
987 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 .handler = icmp_redirect,
989 .error = 1,
990 },
991 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 .handler = icmp_discard,
993 .error = 1,
994 },
995 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 .handler = icmp_discard,
997 .error = 1,
998 },
999 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 .handler = icmp_echo,
1001 },
1002 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 .handler = icmp_discard,
1004 .error = 1,
1005 },
1006 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 .handler = icmp_discard,
1008 .error = 1,
1009 },
1010 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 .handler = icmp_unreach,
1012 .error = 1,
1013 },
1014 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 .handler = icmp_unreach,
1016 .error = 1,
1017 },
1018 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 .handler = icmp_timestamp,
1020 },
1021 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 .handler = icmp_discard,
1023 },
1024 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 .handler = icmp_discard,
1026 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001027 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 .handler = icmp_discard,
1029 },
1030 [ICMP_ADDRESS] = {
David S. Miller838942a2012-07-23 13:20:26 -07001031 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 },
1033 [ICMP_ADDRESSREPLY] = {
David S. Miller838942a2012-07-23 13:20:26 -07001034 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 },
1036};
1037
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001038static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 int i;
1041
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001042 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001043 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001044 kfree(net->ipv4.icmp_sk);
1045 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001046}
1047
Adrian Bunk263173a2008-04-21 02:31:23 -07001048static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001049{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001050 int i, err;
1051
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001052 net->ipv4.icmp_sk =
1053 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1054 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001055 return -ENOMEM;
1056
Denis V. Luneva5710d62008-02-29 11:14:50 -08001057 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001058 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001060 err = inet_ctl_sock_create(&sk, PF_INET,
1061 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001063 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001065 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
1067 /* Enough space for 2 64K ICMP packets, including
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001068 * sk_buff/skb_shared_info struct overhead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 */
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001070 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001072 /*
1073 * Speedup sock_wfree()
1074 */
1075 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001076 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001078
1079 /* Control parameters for ECHO replies. */
1080 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1081 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1082
1083 /* Control parameter - ignore bogus broadcast responses? */
1084 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1085
1086 /*
1087 * Configurable global rate limit.
1088 *
1089 * ratelimit defines tokens/packet consumed for dst->rate_token
1090 * bucket ratemask defines which icmp types are ratelimited by
1091 * setting it's bit position.
1092 *
1093 * default:
1094 * dest unreachable (3), source quench (4),
1095 * time exceeded (11), parameter problem (12)
1096 */
1097
1098 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1099 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1100 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1101
Denis V. Luneva5710d62008-02-29 11:14:50 -08001102 return 0;
1103
1104fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001105 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001106 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001107 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001108 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109}
1110
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001111static struct pernet_operations __net_initdata icmp_sk_ops = {
1112 .init = icmp_sk_init,
1113 .exit = icmp_sk_exit,
1114};
1115
1116int __init icmp_init(void)
1117{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001118 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001119}