blob: 76e10b47e053fd7fc2cbc7fa8d231f54704d1b25 [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 ||
506 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 goto out;
508
509 /*
510 * No replies to physical multicast/broadcast
511 */
512 if (skb_in->pkt_type != PACKET_HOST)
513 goto out;
514
515 /*
516 * Now check at the protocol level
517 */
518 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
519 goto out;
520
521 /*
522 * Only reply to fragment 0. We byte re-order the constant
523 * mask for efficiency.
524 */
525 if (iph->frag_off & htons(IP_OFFSET))
526 goto out;
527
528 /*
529 * If we send an ICMP error to an ICMP error a mess would result..
530 */
531 if (icmp_pointers[type].error) {
532 /*
533 * We are an error, check if we are replying to an
534 * ICMP error
535 */
536 if (iph->protocol == IPPROTO_ICMP) {
537 u8 _inner_type, *itp;
538
539 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700540 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 (iph->ihl << 2) +
542 offsetof(struct icmphdr,
543 type) -
544 skb_in->data,
545 sizeof(_inner_type),
546 &_inner_type);
547 if (itp == NULL)
548 goto out;
549
550 /*
551 * Assume any unknown ICMP type is an error. This
552 * isn't specified by the RFC, but think about it..
553 */
554 if (*itp > NR_ICMP_TYPES ||
555 icmp_pointers[*itp].error)
556 goto out;
557 }
558 }
559
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700560 sk = icmp_xmit_lock(net);
561 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 return;
563
564 /*
565 * Construct source address and options.
566 */
567
568 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700569 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700570 struct net_device *dev = NULL;
571
Eric Dumazet685c7942009-11-01 19:31:03 +0000572 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800573 if (rt_is_input_route(rt) &&
574 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller92101b32012-07-23 16:29:00 -0700575 dev = dev_get_by_index_rcu(net, inet_iif(skb_in));
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700576
Eric Dumazet685c7942009-11-01 19:31:03 +0000577 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700578 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000579 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700580 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000581 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700582 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
584 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
585 IPTOS_PREC_INTERNETCONTROL) :
586 iph->tos;
587
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000588 if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800589 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590
591
592 /*
593 * Prepare data for ICMP header.
594 */
595
596 icmp_param.data.icmph.type = type;
597 icmp_param.data.icmph.code = code;
598 icmp_param.data.icmph.un.gateway = info;
599 icmp_param.data.icmph.checksum = 0;
600 icmp_param.skb = skb_in;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700601 icmp_param.offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800602 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 ipc.addr = iph->saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000604 ipc.opt = &icmp_param.replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000605 ipc.tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
David S. Miller77968b72011-05-08 17:12:19 -0700607 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos,
David S. Millerf6d460c2011-03-01 15:49:55 -0800608 type, code, &icmp_param);
609 if (IS_ERR(rt))
610 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800611
David S. Millera48eff12011-05-18 18:42:43 -0400612 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 goto ende;
614
615 /* RFC says return as much as we can without exceeding 576 bytes. */
616
Changli Gaod8d1f302010-06-10 23:31:35 -0700617 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 if (room > 576)
619 room = 576;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000620 room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 room -= sizeof(struct icmphdr);
622
623 icmp_param.data_len = skb_in->len - icmp_param.offset;
624 if (icmp_param.data_len > room)
625 icmp_param.data_len = room;
626 icmp_param.head_len = sizeof(struct icmphdr);
627
David S. Miller77968b72011-05-08 17:12:19 -0700628 icmp_push_reply(&icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629ende:
630 ip_rt_put(rt);
631out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800632 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633out:;
634}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000635EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
637
David S. Miller1de92432012-07-11 18:32:17 -0700638static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
639{
640 const struct iphdr *iph = (const struct iphdr *) skb->data;
641 const struct net_protocol *ipprot;
642 int protocol = iph->protocol;
643
David S. Millerf0a70e92012-07-12 08:06:04 -0700644 /* Checkin full IP header plus 8 bytes of protocol to
645 * avoid additional coding at protocol handlers.
646 */
647 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
648 return;
649
David S. Miller1de92432012-07-11 18:32:17 -0700650 raw_icmp_error(skb, protocol, info);
651
652 rcu_read_lock();
653 ipprot = rcu_dereference(inet_protos[protocol]);
654 if (ipprot && ipprot->err_handler)
655 ipprot->err_handler(skb, info);
656 rcu_read_unlock();
657}
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659/*
660 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
661 */
662
663static void icmp_unreach(struct sk_buff *skb)
664{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000665 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 struct icmphdr *icmph;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800667 struct net *net;
David S. Millerf9242b62012-06-19 18:56:21 -0700668 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800669
Eric Dumazetadf30902009-06-02 05:19:30 +0000670 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
672 /*
673 * Incomplete header ?
674 * Only checks for the IP header, there should be an
675 * additional check for longer headers in upper levels.
676 */
677
678 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
679 goto out_err;
680
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300681 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000682 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
684 if (iph->ihl < 5) /* Mangled header, drop. */
685 goto out_err;
686
687 if (icmph->type == ICMP_DEST_UNREACH) {
688 switch (icmph->code & 15) {
689 case ICMP_NET_UNREACH:
690 case ICMP_HOST_UNREACH:
691 case ICMP_PROT_UNREACH:
692 case ICMP_PORT_UNREACH:
693 break;
694 case ICMP_FRAG_NEEDED:
695 if (ipv4_config.no_pmtu_disc) {
Joe Perchesafd465032012-03-12 07:03:32 +0000696 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700697 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 } else {
David S. Miller46517002012-06-10 00:04:12 -0700699 info = ntohs(icmph->un.frag.mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 if (!info)
701 goto out;
702 }
703 break;
704 case ICMP_SR_FAILED:
Joe Perchesafd465032012-03-12 07:03:32 +0000705 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700706 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 break;
708 default:
709 break;
710 }
711 if (icmph->code > NR_ICMP_UNREACH)
712 goto out;
713 } else if (icmph->type == ICMP_PARAMETERPROB)
714 info = ntohl(icmph->un.gateway) >> 24;
715
716 /*
717 * Throw it at our lower layers
718 *
719 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
720 * header.
721 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
722 * transport layer.
723 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
724 * transport layer.
725 */
726
727 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300728 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 * bogus responses to broadcast frames. If you see this message
730 * first check your netmask matches at both ends, if it does then
731 * get the other vendor to fix their kit.
732 */
733
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700734 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800735 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Joe Perchese87cc472012-05-13 21:56:26 +0000736 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
737 &ip_hdr(skb)->saddr,
738 icmph->type, icmph->code,
739 &iph->daddr, skb->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 goto out;
741 }
742
David S. Miller1de92432012-07-11 18:32:17 -0700743 icmp_socket_deliver(skb, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744
745out:
746 return;
747out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700748 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 goto out;
750}
751
752
753/*
754 * Handle ICMP_REDIRECT.
755 */
756
757static void icmp_redirect(struct sk_buff *skb)
758{
David S. Miller94206122012-07-11 20:38:08 -0700759 if (skb->len < sizeof(struct iphdr)) {
760 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
761 return;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900762 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000763
David S. Miller94206122012-07-11 20:38:08 -0700764 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
765 return;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000766
David S. Miller94206122012-07-11 20:38:08 -0700767 icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768}
769
770/*
771 * Handle ICMP_ECHO ("ping") requests.
772 *
773 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
774 * requests.
775 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
776 * included in the reply.
777 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
778 * echo requests, MUST have default=NOT.
779 * See also WRT handling of options once they are done and working.
780 */
781
782static void icmp_echo(struct sk_buff *skb)
783{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700784 struct net *net;
785
Eric Dumazetadf30902009-06-02 05:19:30 +0000786 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700787 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 struct icmp_bxm icmp_param;
789
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300790 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
792 icmp_param.skb = skb;
793 icmp_param.offset = 0;
794 icmp_param.data_len = skb->len;
795 icmp_param.head_len = sizeof(struct icmphdr);
796 icmp_reply(&icmp_param, skb);
797 }
798}
799
800/*
801 * Handle ICMP Timestamp requests.
802 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
803 * SHOULD be in the kernel for minimum random latency.
804 * MUST be accurate to a few minutes.
805 * MUST be updated at least at 15Hz.
806 */
807static void icmp_timestamp(struct sk_buff *skb)
808{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700809 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 struct icmp_bxm icmp_param;
811 /*
812 * Too short.
813 */
814 if (skb->len < 4)
815 goto out_err;
816
817 /*
818 * Fill in the current time as ms since midnight UT:
819 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700820 getnstimeofday(&tv);
821 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
822 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 icmp_param.data.times[2] = icmp_param.data.times[1];
824 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
825 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300826 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
828 icmp_param.data.icmph.code = 0;
829 icmp_param.skb = skb;
830 icmp_param.offset = 0;
831 icmp_param.data_len = 0;
832 icmp_param.head_len = sizeof(struct icmphdr) + 12;
833 icmp_reply(&icmp_param, skb);
834out:
835 return;
836out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000837 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 goto out;
839}
840
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841static void icmp_discard(struct sk_buff *skb)
842{
843}
844
845/*
846 * Deal with incoming ICMP packets.
847 */
848int icmp_rcv(struct sk_buff *skb)
849{
850 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000851 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700852 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Herbert Xuaebcf822007-12-12 18:54:16 -0800854 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700855 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800856 int nh;
857
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700858 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800859 XFRM_STATE_ICMP))
860 goto drop;
861
Herbert Xu8b7817f2007-12-12 10:44:43 -0800862 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
863 goto drop;
864
865 nh = skb_network_offset(skb);
866 skb_set_network_header(skb, sizeof(*icmph));
867
868 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
869 goto drop;
870
871 skb_set_network_header(skb, nh);
872 }
873
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700874 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875
876 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700877 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800878 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 break;
Herbert Xufb286bb2005-11-10 13:01:24 -0800880 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -0800882 skb->csum = 0;
883 if (__skb_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000884 goto csum_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 }
886
Herbert Xu8cf22942008-02-05 03:15:50 -0800887 if (!pskb_pull(skb, sizeof(*icmph)))
888 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300890 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
Pavel Emelyanovf66ac032008-07-14 23:04:00 -0700892 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 /*
894 * 18 is the highest 'known' ICMP type. Anything else is a mystery
895 *
896 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
897 * discarded.
898 */
899 if (icmph->type > NR_ICMP_TYPES)
900 goto error;
901
902
903 /*
904 * Parse the ICMP message
905 */
906
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900907 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 /*
909 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
910 * silently ignored (we let user decide with a sysctl).
911 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
912 * discarded if to broadcast/multicast.
913 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -0700914 if ((icmph->type == ICMP_ECHO ||
915 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700916 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 goto error;
918 }
919 if (icmph->type != ICMP_ECHO &&
920 icmph->type != ICMP_TIMESTAMP &&
921 icmph->type != ICMP_ADDRESS &&
922 icmph->type != ICMP_ADDRESSREPLY) {
923 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900924 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 }
926
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 icmp_pointers[icmph->type].handler(skb);
928
929drop:
930 kfree_skb(skb);
931 return 0;
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000932csum_error:
933 ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700935 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 goto drop;
937}
938
Li Wei5b052042013-02-21 22:18:44 +0000939void icmp_err(struct sk_buff *skb, u32 info)
940{
941 struct iphdr *iph = (struct iphdr *)skb->data;
942 struct icmphdr *icmph = (struct icmphdr *)(skb->data+(iph->ihl<<2));
943 int type = icmp_hdr(skb)->type;
944 int code = icmp_hdr(skb)->code;
945 struct net *net = dev_net(skb->dev);
946
947 /*
948 * Use ping_err to handle all icmp errors except those
949 * triggered by ICMP_ECHOREPLY which sent from kernel.
950 */
951 if (icmph->type != ICMP_ECHOREPLY) {
952 ping_err(skb, info);
953 return;
954 }
955
956 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
957 ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0);
958 else if (type == ICMP_REDIRECT)
959 ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0);
960}
961
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962/*
963 * This table is the definition of how we handle ICMP.
964 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800965static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000967 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 },
969 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 .handler = icmp_discard,
971 .error = 1,
972 },
973 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 .handler = icmp_discard,
975 .error = 1,
976 },
977 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 .handler = icmp_unreach,
979 .error = 1,
980 },
981 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 .handler = icmp_unreach,
983 .error = 1,
984 },
985 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 .handler = icmp_redirect,
987 .error = 1,
988 },
989 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 .handler = icmp_discard,
991 .error = 1,
992 },
993 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 .handler = icmp_discard,
995 .error = 1,
996 },
997 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 .handler = icmp_echo,
999 },
1000 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 .handler = icmp_discard,
1002 .error = 1,
1003 },
1004 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 .handler = icmp_discard,
1006 .error = 1,
1007 },
1008 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 .handler = icmp_unreach,
1010 .error = 1,
1011 },
1012 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 .handler = icmp_unreach,
1014 .error = 1,
1015 },
1016 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 .handler = icmp_timestamp,
1018 },
1019 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 .handler = icmp_discard,
1021 },
1022 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 .handler = icmp_discard,
1024 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001025 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 .handler = icmp_discard,
1027 },
1028 [ICMP_ADDRESS] = {
David S. Miller838942a2012-07-23 13:20:26 -07001029 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 },
1031 [ICMP_ADDRESSREPLY] = {
David S. Miller838942a2012-07-23 13:20:26 -07001032 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 },
1034};
1035
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001036static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 int i;
1039
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001040 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001041 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001042 kfree(net->ipv4.icmp_sk);
1043 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001044}
1045
Adrian Bunk263173a2008-04-21 02:31:23 -07001046static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001047{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001048 int i, err;
1049
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001050 net->ipv4.icmp_sk =
1051 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1052 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001053 return -ENOMEM;
1054
Denis V. Luneva5710d62008-02-29 11:14:50 -08001055 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001056 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001058 err = inet_ctl_sock_create(&sk, PF_INET,
1059 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001061 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001063 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064
1065 /* Enough space for 2 64K ICMP packets, including
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001066 * sk_buff/skb_shared_info struct overhead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 */
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001068 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001070 /*
1071 * Speedup sock_wfree()
1072 */
1073 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001074 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001076
1077 /* Control parameters for ECHO replies. */
1078 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1079 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1080
1081 /* Control parameter - ignore bogus broadcast responses? */
1082 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1083
1084 /*
1085 * Configurable global rate limit.
1086 *
1087 * ratelimit defines tokens/packet consumed for dst->rate_token
1088 * bucket ratemask defines which icmp types are ratelimited by
1089 * setting it's bit position.
1090 *
1091 * default:
1092 * dest unreachable (3), source quench (4),
1093 * time exceeded (11), parameter problem (12)
1094 */
1095
1096 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1097 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1098 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1099
Denis V. Luneva5710d62008-02-29 11:14:50 -08001100 return 0;
1101
1102fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001103 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001104 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001105 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001106 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107}
1108
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001109static struct pernet_operations __net_initdata icmp_sk_ops = {
1110 .init = icmp_sk_init,
1111 .exit = icmp_sk_exit,
1112};
1113
1114int __init icmp_init(void)
1115{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001116 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001117}