blob: 5c0e8bc6e5ba275d2469336533b5848de94ae6c1 [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;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200356 ipc.ttl = 0;
357 ipc.tos = -1;
358
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000359 if (icmp_param->replyopts.opt.opt.optlen) {
360 ipc.opt = &icmp_param->replyopts.opt;
361 if (ipc.opt->opt.srr)
362 daddr = icmp_param->replyopts.opt.opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 }
David S. Miller77968b72011-05-08 17:12:19 -0700364 memset(&fl4, 0, sizeof(fl4));
365 fl4.daddr = daddr;
David S. Miller35ebf652012-06-28 03:59:11 -0700366 fl4.saddr = saddr;
David S. Miller77968b72011-05-08 17:12:19 -0700367 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
368 fl4.flowi4_proto = IPPROTO_ICMP;
369 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
370 rt = ip_route_output_key(net, &fl4);
371 if (IS_ERR(rt))
372 goto out_unlock;
David S. Millera48eff12011-05-18 18:42:43 -0400373 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 icmp_param->data.icmph.code))
David S. Miller77968b72011-05-08 17:12:19 -0700375 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 ip_rt_put(rt);
377out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800378 icmp_xmit_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379}
380
David S. Miller77968b72011-05-08 17:12:19 -0700381static struct rtable *icmp_route_lookup(struct net *net,
382 struct flowi4 *fl4,
383 struct sk_buff *skb_in,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000384 const struct iphdr *iph,
David S. Millerf6d460c2011-03-01 15:49:55 -0800385 __be32 saddr, u8 tos,
386 int type, int code,
387 struct icmp_bxm *param)
388{
David S. Millerf6d460c2011-03-01 15:49:55 -0800389 struct rtable *rt, *rt2;
David S. Miller415b3332011-07-22 06:22:10 -0700390 struct flowi4 fl4_dec;
David S. Millerf6d460c2011-03-01 15:49:55 -0800391 int err;
392
David S. Miller77968b72011-05-08 17:12:19 -0700393 memset(fl4, 0, sizeof(*fl4));
394 fl4->daddr = (param->replyopts.opt.opt.srr ?
395 param->replyopts.opt.opt.faddr : iph->saddr);
396 fl4->saddr = saddr;
397 fl4->flowi4_tos = RT_TOS(tos);
398 fl4->flowi4_proto = IPPROTO_ICMP;
399 fl4->fl4_icmp_type = type;
400 fl4->fl4_icmp_code = code;
401 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
402 rt = __ip_route_output_key(net, fl4);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800403 if (IS_ERR(rt))
404 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800405
406 /* No need to clone since we're just using its address. */
407 rt2 = rt;
408
David S. Miller9d6ec932011-03-12 01:12:47 -0500409 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
David S. Miller77968b72011-05-08 17:12:19 -0700410 flowi4_to_flowi(fl4), NULL, 0);
David S. Miller452edd52011-03-02 13:27:41 -0800411 if (!IS_ERR(rt)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800412 if (rt != rt2)
413 return rt;
David S. Miller452edd52011-03-02 13:27:41 -0800414 } else if (PTR_ERR(rt) == -EPERM) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800415 rt = NULL;
David S. Miller452edd52011-03-02 13:27:41 -0800416 } else
417 return rt;
David S. Millerf6d460c2011-03-01 15:49:55 -0800418
David S. Miller415b3332011-07-22 06:22:10 -0700419 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
David S. Millerf6d460c2011-03-01 15:49:55 -0800420 if (err)
421 goto relookup_failed;
422
David S. Miller415b3332011-07-22 06:22:10 -0700423 if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) {
424 rt2 = __ip_route_output_key(net, &fl4_dec);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800425 if (IS_ERR(rt2))
426 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800427 } else {
David S. Miller9d6ec932011-03-12 01:12:47 -0500428 struct flowi4 fl4_2 = {};
David S. Millerf6d460c2011-03-01 15:49:55 -0800429 unsigned long orefdst;
430
David S. Miller415b3332011-07-22 06:22:10 -0700431 fl4_2.daddr = fl4_dec.saddr;
David S. Miller9d6ec932011-03-12 01:12:47 -0500432 rt2 = ip_route_output_key(net, &fl4_2);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800433 if (IS_ERR(rt2)) {
434 err = PTR_ERR(rt2);
David S. Millerf6d460c2011-03-01 15:49:55 -0800435 goto relookup_failed;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800436 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800437 /* Ugh! */
438 orefdst = skb_in->_skb_refdst; /* save old refdst */
David S. Miller415b3332011-07-22 06:22:10 -0700439 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
David S. Millerf6d460c2011-03-01 15:49:55 -0800440 RT_TOS(tos), rt2->dst.dev);
441
442 dst_release(&rt2->dst);
443 rt2 = skb_rtable(skb_in);
444 skb_in->_skb_refdst = orefdst; /* restore old refdst */
445 }
446
447 if (err)
448 goto relookup_failed;
449
David S. Miller9d6ec932011-03-12 01:12:47 -0500450 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
David S. Miller415b3332011-07-22 06:22:10 -0700451 flowi4_to_flowi(&fl4_dec), NULL,
David S. Miller9d6ec932011-03-12 01:12:47 -0500452 XFRM_LOOKUP_ICMP);
David S. Miller452edd52011-03-02 13:27:41 -0800453 if (!IS_ERR(rt2)) {
David S. Millerf6d460c2011-03-01 15:49:55 -0800454 dst_release(&rt->dst);
David S. Miller415b3332011-07-22 06:22:10 -0700455 memcpy(fl4, &fl4_dec, sizeof(*fl4));
David S. Millerf6d460c2011-03-01 15:49:55 -0800456 rt = rt2;
David S. Miller452edd52011-03-02 13:27:41 -0800457 } else if (PTR_ERR(rt2) == -EPERM) {
458 if (rt)
459 dst_release(&rt->dst);
460 return rt2;
461 } else {
462 err = PTR_ERR(rt2);
463 goto relookup_failed;
David S. Millerf6d460c2011-03-01 15:49:55 -0800464 }
David S. Millerf6d460c2011-03-01 15:49:55 -0800465 return rt;
466
467relookup_failed:
468 if (rt)
469 return rt;
470 return ERR_PTR(err);
471}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
473/*
474 * Send an ICMP message in response to a situation
475 *
476 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
477 * MAY send more (we do).
478 * MUST NOT change this header information.
479 * MUST NOT reply to a multicast/broadcast IP address.
480 * MUST NOT reply to a multicast/broadcast MAC address.
481 * MUST reply to only the first fragment.
482 */
483
Al Viroe4883012006-09-26 22:17:28 -0700484void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485{
486 struct iphdr *iph;
487 int room;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000488 struct icmp_bxm *icmp_param;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000489 struct rtable *rt = skb_rtable(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 struct ipcm_cookie ipc;
David S. Miller77968b72011-05-08 17:12:19 -0700491 struct flowi4 fl4;
Al Viroa61ced52006-09-26 21:27:54 -0700492 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 u8 tos;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800494 struct net *net;
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -0800495 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
497 if (!rt)
498 goto out;
Changli Gaod8d1f302010-06-10 23:31:35 -0700499 net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500
501 /*
502 * Find the original header. It is expected to be valid, of course.
503 * Check this, icmp_send is called from the most obscure devices
504 * sometimes.
505 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700506 iph = ip_hdr(skb_in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700508 if ((u8 *)iph < skb_in->head ||
Simon Hormanf7c0c2a2013-05-28 20:34:27 +0000509 (skb_network_header(skb_in) + sizeof(*iph)) >
510 skb_tail_pointer(skb_in))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 goto out;
512
513 /*
514 * No replies to physical multicast/broadcast
515 */
516 if (skb_in->pkt_type != PACKET_HOST)
517 goto out;
518
519 /*
520 * Now check at the protocol level
521 */
522 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
523 goto out;
524
525 /*
526 * Only reply to fragment 0. We byte re-order the constant
527 * mask for efficiency.
528 */
529 if (iph->frag_off & htons(IP_OFFSET))
530 goto out;
531
532 /*
533 * If we send an ICMP error to an ICMP error a mess would result..
534 */
535 if (icmp_pointers[type].error) {
536 /*
537 * We are an error, check if we are replying to an
538 * ICMP error
539 */
540 if (iph->protocol == IPPROTO_ICMP) {
541 u8 _inner_type, *itp;
542
543 itp = skb_header_pointer(skb_in,
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700544 skb_network_header(skb_in) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 (iph->ihl << 2) +
546 offsetof(struct icmphdr,
547 type) -
548 skb_in->data,
549 sizeof(_inner_type),
550 &_inner_type);
551 if (itp == NULL)
552 goto out;
553
554 /*
555 * Assume any unknown ICMP type is an error. This
556 * isn't specified by the RFC, but think about it..
557 */
558 if (*itp > NR_ICMP_TYPES ||
559 icmp_pointers[*itp].error)
560 goto out;
561 }
562 }
563
Cong Wang9a99d4a2013-06-02 15:00:52 +0000564 icmp_param = kmalloc(sizeof(*icmp_param), GFP_ATOMIC);
565 if (!icmp_param)
566 return;
567
Denis V. Lunevfdc0bde2008-08-23 04:43:33 -0700568 sk = icmp_xmit_lock(net);
569 if (sk == NULL)
Cong Wang9a99d4a2013-06-02 15:00:52 +0000570 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572 /*
573 * Construct source address and options.
574 */
575
576 saddr = iph->daddr;
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700577 if (!(rt->rt_flags & RTCF_LOCAL)) {
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700578 struct net_device *dev = NULL;
579
Eric Dumazet685c7942009-11-01 19:31:03 +0000580 rcu_read_lock();
David S. Millerc7537962010-11-11 17:07:48 -0800581 if (rt_is_input_route(rt) &&
582 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
David S. Miller92101b32012-07-23 16:29:00 -0700583 dev = dev_get_by_index_rcu(net, inet_iif(skb_in));
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700584
Eric Dumazet685c7942009-11-01 19:31:03 +0000585 if (dev)
Patrick McHardy6e1d9102007-06-01 11:45:04 -0700586 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
Eric Dumazet685c7942009-11-01 19:31:03 +0000587 else
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700588 saddr = 0;
Eric Dumazet685c7942009-11-01 19:31:03 +0000589 rcu_read_unlock();
J. Simonetti1c2fb7f2005-06-13 15:19:03 -0700590 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591
592 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
593 IPTOS_PREC_INTERNETCONTROL) :
594 iph->tos;
595
Cong Wang9a99d4a2013-06-02 15:00:52 +0000596 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb_in))
Herbert Xufa60cf72006-02-04 02:09:34 -0800597 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
599
600 /*
601 * Prepare data for ICMP header.
602 */
603
Cong Wang9a99d4a2013-06-02 15:00:52 +0000604 icmp_param->data.icmph.type = type;
605 icmp_param->data.icmph.code = code;
606 icmp_param->data.icmph.un.gateway = info;
607 icmp_param->data.icmph.checksum = 0;
608 icmp_param->skb = skb_in;
609 icmp_param->offset = skb_network_offset(skb_in);
Denis V. Lunev405666d2008-02-29 11:16:46 -0800610 inet_sk(sk)->tos = tos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 ipc.addr = iph->saddr;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000612 ipc.opt = &icmp_param->replyopts.opt;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000613 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200614 ipc.ttl = 0;
615 ipc.tos = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
David S. Miller77968b72011-05-08 17:12:19 -0700617 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos,
Cong Wang9a99d4a2013-06-02 15:00:52 +0000618 type, code, icmp_param);
David S. Millerf6d460c2011-03-01 15:49:55 -0800619 if (IS_ERR(rt))
620 goto out_unlock;
Herbert Xu8b7817f2007-12-12 10:44:43 -0800621
David S. Millera48eff12011-05-18 18:42:43 -0400622 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 goto ende;
624
625 /* RFC says return as much as we can without exceeding 576 bytes. */
626
Changli Gaod8d1f302010-06-10 23:31:35 -0700627 room = dst_mtu(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 if (room > 576)
629 room = 576;
Cong Wang9a99d4a2013-06-02 15:00:52 +0000630 room -= sizeof(struct iphdr) + icmp_param->replyopts.opt.opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 room -= sizeof(struct icmphdr);
632
Cong Wang9a99d4a2013-06-02 15:00:52 +0000633 icmp_param->data_len = skb_in->len - icmp_param->offset;
634 if (icmp_param->data_len > room)
635 icmp_param->data_len = room;
636 icmp_param->head_len = sizeof(struct icmphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637
Cong Wang9a99d4a2013-06-02 15:00:52 +0000638 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639ende:
640 ip_rt_put(rt);
641out_unlock:
Denis V. Lunev405666d2008-02-29 11:16:46 -0800642 icmp_xmit_unlock(sk);
Cong Wang9a99d4a2013-06-02 15:00:52 +0000643out_free:
644 kfree(icmp_param);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645out:;
646}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000647EXPORT_SYMBOL(icmp_send);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
649
David S. Miller1de92432012-07-11 18:32:17 -0700650static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
651{
652 const struct iphdr *iph = (const struct iphdr *) skb->data;
653 const struct net_protocol *ipprot;
654 int protocol = iph->protocol;
655
David S. Millerf0a70e92012-07-12 08:06:04 -0700656 /* Checkin full IP header plus 8 bytes of protocol to
657 * avoid additional coding at protocol handlers.
658 */
659 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
660 return;
661
David S. Miller1de92432012-07-11 18:32:17 -0700662 raw_icmp_error(skb, protocol, info);
663
664 rcu_read_lock();
665 ipprot = rcu_dereference(inet_protos[protocol]);
666 if (ipprot && ipprot->err_handler)
667 ipprot->err_handler(skb, info);
668 rcu_read_unlock();
669}
670
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671/*
Rami Rosen08578d82013-06-03 00:23:25 +0000672 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, ICMP_QUENCH, and
673 * ICMP_PARAMETERPROB.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 */
675
676static void icmp_unreach(struct sk_buff *skb)
677{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000678 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 struct icmphdr *icmph;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800680 struct net *net;
David S. Millerf9242b62012-06-19 18:56:21 -0700681 u32 info = 0;
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800682
Eric Dumazetadf30902009-06-02 05:19:30 +0000683 net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
685 /*
686 * Incomplete header ?
687 * Only checks for the IP header, there should be an
688 * additional check for longer headers in upper levels.
689 */
690
691 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
692 goto out_err;
693
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300694 icmph = icmp_hdr(skb);
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000695 iph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
697 if (iph->ihl < 5) /* Mangled header, drop. */
698 goto out_err;
699
700 if (icmph->type == ICMP_DEST_UNREACH) {
701 switch (icmph->code & 15) {
702 case ICMP_NET_UNREACH:
703 case ICMP_HOST_UNREACH:
704 case ICMP_PROT_UNREACH:
705 case ICMP_PORT_UNREACH:
706 break;
707 case ICMP_FRAG_NEEDED:
708 if (ipv4_config.no_pmtu_disc) {
Joe Perchesafd465032012-03-12 07:03:32 +0000709 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700710 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 } else {
David S. Miller46517002012-06-10 00:04:12 -0700712 info = ntohs(icmph->un.frag.mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 if (!info)
714 goto out;
715 }
716 break;
717 case ICMP_SR_FAILED:
Joe Perchesafd465032012-03-12 07:03:32 +0000718 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"),
Harvey Harrison673d57e2008-10-31 00:53:57 -0700719 &iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 break;
721 default:
722 break;
723 }
724 if (icmph->code > NR_ICMP_UNREACH)
725 goto out;
726 } else if (icmph->type == ICMP_PARAMETERPROB)
727 info = ntohl(icmph->un.gateway) >> 24;
728
729 /*
730 * Throw it at our lower layers
731 *
732 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
733 * header.
734 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
735 * transport layer.
736 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
737 * transport layer.
738 */
739
740 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300741 * Check the other end isn't violating RFC 1122. Some routers send
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 * bogus responses to broadcast frames. If you see this message
743 * first check your netmask matches at both ends, if it does then
744 * get the other vendor to fix their kit.
745 */
746
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700747 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
Denis V. Lunevdde1bc02008-01-22 23:50:57 -0800748 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
Joe Perchese87cc472012-05-13 21:56:26 +0000749 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
750 &ip_hdr(skb)->saddr,
751 icmph->type, icmph->code,
752 &iph->daddr, skb->dev->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 goto out;
754 }
755
David S. Miller1de92432012-07-11 18:32:17 -0700756 icmp_socket_deliver(skb, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
758out:
759 return;
760out_err:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700761 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 goto out;
763}
764
765
766/*
767 * Handle ICMP_REDIRECT.
768 */
769
770static void icmp_redirect(struct sk_buff *skb)
771{
David S. Miller94206122012-07-11 20:38:08 -0700772 if (skb->len < sizeof(struct iphdr)) {
773 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
774 return;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900775 }
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000776
David S. Miller94206122012-07-11 20:38:08 -0700777 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
778 return;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000779
David S. Miller94206122012-07-11 20:38:08 -0700780 icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
782
783/*
784 * Handle ICMP_ECHO ("ping") requests.
785 *
786 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
787 * requests.
788 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
789 * included in the reply.
790 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
791 * echo requests, MUST have default=NOT.
792 * See also WRT handling of options once they are done and working.
793 */
794
795static void icmp_echo(struct sk_buff *skb)
796{
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700797 struct net *net;
798
Eric Dumazetadf30902009-06-02 05:19:30 +0000799 net = dev_net(skb_dst(skb)->dev);
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700800 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 struct icmp_bxm icmp_param;
802
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300803 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
805 icmp_param.skb = skb;
806 icmp_param.offset = 0;
807 icmp_param.data_len = skb->len;
808 icmp_param.head_len = sizeof(struct icmphdr);
809 icmp_reply(&icmp_param, skb);
810 }
811}
812
813/*
814 * Handle ICMP Timestamp requests.
815 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
816 * SHOULD be in the kernel for minimum random latency.
817 * MUST be accurate to a few minutes.
818 * MUST be updated at least at 15Hz.
819 */
820static void icmp_timestamp(struct sk_buff *skb)
821{
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700822 struct timespec tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 struct icmp_bxm icmp_param;
824 /*
825 * Too short.
826 */
827 if (skb->len < 4)
828 goto out_err;
829
830 /*
831 * Fill in the current time as ms since midnight UT:
832 */
YOSHIFUJI Hideakif25c3d62008-04-21 02:34:08 -0700833 getnstimeofday(&tv);
834 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
835 tv.tv_nsec / NSEC_PER_MSEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 icmp_param.data.times[2] = icmp_param.data.times[1];
837 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
838 BUG();
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300839 icmp_param.data.icmph = *icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
841 icmp_param.data.icmph.code = 0;
842 icmp_param.skb = skb;
843 icmp_param.offset = 0;
844 icmp_param.data_len = 0;
845 icmp_param.head_len = sizeof(struct icmphdr) + 12;
846 icmp_reply(&icmp_param, skb);
847out:
848 return;
849out_err:
Eric Dumazetadf30902009-06-02 05:19:30 +0000850 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 goto out;
852}
853
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854static void icmp_discard(struct sk_buff *skb)
855{
856}
857
858/*
859 * Deal with incoming ICMP packets.
860 */
861int icmp_rcv(struct sk_buff *skb)
862{
863 struct icmphdr *icmph;
Eric Dumazet511c3f92009-06-02 05:14:27 +0000864 struct rtable *rt = skb_rtable(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700865 struct net *net = dev_net(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
Herbert Xuaebcf822007-12-12 18:54:16 -0800867 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700868 struct sec_path *sp = skb_sec_path(skb);
Herbert Xu8b7817f2007-12-12 10:44:43 -0800869 int nh;
870
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700871 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
Herbert Xuaebcf822007-12-12 18:54:16 -0800872 XFRM_STATE_ICMP))
873 goto drop;
874
Herbert Xu8b7817f2007-12-12 10:44:43 -0800875 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
876 goto drop;
877
878 nh = skb_network_offset(skb);
879 skb_set_network_header(skb, sizeof(*icmph));
880
881 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
882 goto drop;
883
884 skb_set_network_header(skb, nh);
885 }
886
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700887 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888
889 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700890 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800891 if (!csum_fold(skb->csum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 break;
Herbert Xufb286bb2005-11-10 13:01:24 -0800893 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 case CHECKSUM_NONE:
Herbert Xufb286bb2005-11-10 13:01:24 -0800895 skb->csum = 0;
896 if (__skb_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000897 goto csum_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 }
899
Herbert Xu8cf22942008-02-05 03:15:50 -0800900 if (!pskb_pull(skb, sizeof(*icmph)))
901 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300903 icmph = icmp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
Pavel Emelyanovf66ac032008-07-14 23:04:00 -0700905 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 /*
907 * 18 is the highest 'known' ICMP type. Anything else is a mystery
908 *
909 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
910 * discarded.
911 */
912 if (icmph->type > NR_ICMP_TYPES)
913 goto error;
914
915
916 /*
917 * Parse the ICMP message
918 */
919
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900920 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 /*
922 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
923 * silently ignored (we let user decide with a sysctl).
924 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
925 * discarded if to broadcast/multicast.
926 */
Alexey Kuznetsov4c866aa2005-07-08 17:34:46 -0700927 if ((icmph->type == ICMP_ECHO ||
928 icmph->type == ICMP_TIMESTAMP) &&
Pavel Emelyanovb34a95e2008-03-26 02:00:21 -0700929 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 goto error;
931 }
932 if (icmph->type != ICMP_ECHO &&
933 icmph->type != ICMP_TIMESTAMP &&
934 icmph->type != ICMP_ADDRESS &&
935 icmph->type != ICMP_ADDRESSREPLY) {
936 goto error;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900937 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 }
939
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 icmp_pointers[icmph->type].handler(skb);
941
942drop:
943 kfree_skb(skb);
944 return 0;
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000945csum_error:
946 ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947error:
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700948 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 goto drop;
950}
951
Li Wei5b052042013-02-21 22:18:44 +0000952void icmp_err(struct sk_buff *skb, u32 info)
953{
954 struct iphdr *iph = (struct iphdr *)skb->data;
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +0000955 int offset = iph->ihl<<2;
956 struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset);
Li Wei5b052042013-02-21 22:18:44 +0000957 int type = icmp_hdr(skb)->type;
958 int code = icmp_hdr(skb)->code;
959 struct net *net = dev_net(skb->dev);
960
961 /*
962 * Use ping_err to handle all icmp errors except those
963 * triggered by ICMP_ECHOREPLY which sent from kernel.
964 */
965 if (icmph->type != ICMP_ECHOREPLY) {
Lorenzo Colitti6d0bfe22013-05-22 20:17:31 +0000966 ping_err(skb, offset, info);
Li Wei5b052042013-02-21 22:18:44 +0000967 return;
968 }
969
970 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
971 ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0);
972 else if (type == ICMP_REDIRECT)
973 ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0);
974}
975
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976/*
977 * This table is the definition of how we handle ICMP.
978 */
Arjan van de Ven9b5b5cf2005-11-29 16:21:38 -0800979static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 [ICMP_ECHOREPLY] = {
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000981 .handler = ping_rcv,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 },
983 [1] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 .handler = icmp_discard,
985 .error = 1,
986 },
987 [2] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 .handler = icmp_discard,
989 .error = 1,
990 },
991 [ICMP_DEST_UNREACH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 .handler = icmp_unreach,
993 .error = 1,
994 },
995 [ICMP_SOURCE_QUENCH] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 .handler = icmp_unreach,
997 .error = 1,
998 },
999 [ICMP_REDIRECT] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 .handler = icmp_redirect,
1001 .error = 1,
1002 },
1003 [6] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 .handler = icmp_discard,
1005 .error = 1,
1006 },
1007 [7] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 .handler = icmp_discard,
1009 .error = 1,
1010 },
1011 [ICMP_ECHO] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 .handler = icmp_echo,
1013 },
1014 [9] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 .handler = icmp_discard,
1016 .error = 1,
1017 },
1018 [10] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 .handler = icmp_discard,
1020 .error = 1,
1021 },
1022 [ICMP_TIME_EXCEEDED] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 .handler = icmp_unreach,
1024 .error = 1,
1025 },
1026 [ICMP_PARAMETERPROB] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 .handler = icmp_unreach,
1028 .error = 1,
1029 },
1030 [ICMP_TIMESTAMP] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 .handler = icmp_timestamp,
1032 },
1033 [ICMP_TIMESTAMPREPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 .handler = icmp_discard,
1035 },
1036 [ICMP_INFO_REQUEST] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 .handler = icmp_discard,
1038 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001039 [ICMP_INFO_REPLY] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 .handler = icmp_discard,
1041 },
1042 [ICMP_ADDRESS] = {
David S. Miller838942a2012-07-23 13:20:26 -07001043 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 },
1045 [ICMP_ADDRESSREPLY] = {
David S. Miller838942a2012-07-23 13:20:26 -07001046 .handler = icmp_discard,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 },
1048};
1049
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001050static void __net_exit icmp_sk_exit(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 int i;
1053
Denis V. Lunev5c8cafd2008-02-29 11:19:22 -08001054 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001055 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001056 kfree(net->ipv4.icmp_sk);
1057 net->ipv4.icmp_sk = NULL;
Denis V. Luneva5710d62008-02-29 11:14:50 -08001058}
1059
Adrian Bunk263173a2008-04-21 02:31:23 -07001060static int __net_init icmp_sk_init(struct net *net)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001061{
Denis V. Luneva5710d62008-02-29 11:14:50 -08001062 int i, err;
1063
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001064 net->ipv4.icmp_sk =
1065 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1066 if (net->ipv4.icmp_sk == NULL)
Denis V. Lunev79c91152008-02-29 11:17:11 -08001067 return -ENOMEM;
1068
Denis V. Luneva5710d62008-02-29 11:14:50 -08001069 for_each_possible_cpu(i) {
Denis V. Lunev1e3cf682008-02-29 11:15:42 -08001070 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001072 err = inet_ctl_sock_create(&sk, PF_INET,
1073 SOCK_RAW, IPPROTO_ICMP, net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 if (err < 0)
Denis V. Luneva5710d62008-02-29 11:14:50 -08001075 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001077 net->ipv4.icmp_sk[i] = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
1079 /* Enough space for 2 64K ICMP packets, including
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001080 * sk_buff/skb_shared_info struct overhead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 */
Eric Dumazet87fb4b72011-10-13 07:28:54 +00001082 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
Eric Dumazetb3a5b6c2009-09-24 12:16:51 +00001084 /*
1085 * Speedup sock_wfree()
1086 */
1087 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001088 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 }
Pavel Emelyanova24022e2008-03-26 01:55:37 -07001090
1091 /* Control parameters for ECHO replies. */
1092 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1093 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1094
1095 /* Control parameter - ignore bogus broadcast responses? */
1096 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1097
1098 /*
1099 * Configurable global rate limit.
1100 *
1101 * ratelimit defines tokens/packet consumed for dst->rate_token
1102 * bucket ratemask defines which icmp types are ratelimited by
1103 * setting it's bit position.
1104 *
1105 * default:
1106 * dest unreachable (3), source quench (4),
1107 * time exceeded (11), parameter problem (12)
1108 */
1109
1110 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1111 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1112 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1113
Denis V. Luneva5710d62008-02-29 11:14:50 -08001114 return 0;
1115
1116fail:
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001117 for_each_possible_cpu(i)
Denis V. Lunevc1e98942008-04-03 14:29:00 -07001118 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
Denis V. Lunev1d1c8d12008-02-29 14:15:19 -08001119 kfree(net->ipv4.icmp_sk);
Denis V. Luneva5710d62008-02-29 11:14:50 -08001120 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121}
1122
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001123static struct pernet_operations __net_initdata icmp_sk_ops = {
1124 .init = icmp_sk_init,
1125 .exit = icmp_sk_exit,
1126};
1127
1128int __init icmp_init(void)
1129{
Eric W. Biederman959d2722009-02-22 00:09:14 -08001130 return register_pernet_subsys(&icmp_sk_ops);
Denis V. Lunev4a6ad7a2008-02-29 11:19:58 -08001131}