Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NET3: Implementation of the ICMP protocol layer. | |
3 | * | |
113aa838 | 4 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * |
1da177e4 LT |
6 | * 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 | ||
afd46503 JP |
65 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
66 | ||
1da177e4 LT |
67 | #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> | |
14c85021 | 75 | #include <linux/inetdevice.h> |
1da177e4 LT |
76 | #include <linux/netdevice.h> |
77 | #include <linux/string.h> | |
78 | #include <linux/netfilter_ipv4.h> | |
5a0e3ad6 | 79 | #include <linux/slab.h> |
1da177e4 LT |
80 | #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> | |
c319b4d7 | 88 | #include <net/ping.h> |
1da177e4 LT |
89 | #include <linux/skbuff.h> |
90 | #include <net/sock.h> | |
91 | #include <linux/errno.h> | |
92 | #include <linux/timer.h> | |
93 | #include <linux/init.h> | |
1da177e4 LT |
94 | #include <asm/uaccess.h> |
95 | #include <net/checksum.h> | |
8b7817f3 | 96 | #include <net/xfrm.h> |
c1e9894d | 97 | #include <net/inet_common.h> |
35ebf65e | 98 | #include <net/ip_fib.h> |
1da177e4 LT |
99 | |
100 | /* | |
101 | * Build xmit assembly blocks | |
102 | */ | |
103 | ||
104 | struct icmp_bxm { | |
105 | struct sk_buff *skb; | |
106 | int offset; | |
107 | int data_len; | |
108 | ||
109 | struct { | |
110 | struct icmphdr icmph; | |
b03d73e3 | 111 | __be32 times[3]; |
1da177e4 LT |
112 | } data; |
113 | int head_len; | |
f6d8bd05 | 114 | struct ip_options_data replyopts; |
1da177e4 LT |
115 | }; |
116 | ||
1da177e4 LT |
117 | /* 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 | ||
e754834e | 120 | const struct icmp_err icmp_err_convert[] = { |
1da177e4 LT |
121 | { |
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 | }; | |
4bc2f18b | 186 | EXPORT_SYMBOL(icmp_err_convert); |
1da177e4 | 187 | |
1da177e4 LT |
188 | /* |
189 | * ICMP control array. This specifies what to do with each ICMP. | |
190 | */ | |
191 | ||
192 | struct icmp_control { | |
1da177e4 LT |
193 | void (*handler)(struct sk_buff *skb); |
194 | short error; /* This ICMP is classed as an error message */ | |
195 | }; | |
196 | ||
9b5b5cff | 197 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1]; |
1da177e4 LT |
198 | |
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 | */ | |
4a6ad7a1 DL |
206 | static struct sock *icmp_sk(struct net *net) |
207 | { | |
208 | return net->ipv4.icmp_sk[smp_processor_id()]; | |
209 | } | |
1da177e4 | 210 | |
fdc0bde9 | 211 | static inline struct sock *icmp_xmit_lock(struct net *net) |
1da177e4 | 212 | { |
fdc0bde9 DL |
213 | struct sock *sk; |
214 | ||
1da177e4 LT |
215 | local_bh_disable(); |
216 | ||
fdc0bde9 DL |
217 | sk = icmp_sk(net); |
218 | ||
405666db | 219 | if (unlikely(!spin_trylock(&sk->sk_lock.slock))) { |
1da177e4 LT |
220 | /* This can happen if the output path signals a |
221 | * dst_link_failure() for an outgoing ICMP packet. | |
222 | */ | |
223 | local_bh_enable(); | |
fdc0bde9 | 224 | return NULL; |
1da177e4 | 225 | } |
fdc0bde9 | 226 | return sk; |
1da177e4 LT |
227 | } |
228 | ||
405666db | 229 | static inline void icmp_xmit_unlock(struct sock *sk) |
1da177e4 | 230 | { |
405666db | 231 | spin_unlock_bh(&sk->sk_lock.slock); |
1da177e4 LT |
232 | } |
233 | ||
234 | /* | |
235 | * Send an ICMP frame. | |
236 | */ | |
237 | ||
92d86829 | 238 | static inline bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt, |
a48eff12 | 239 | struct flowi4 *fl4, int type, int code) |
1da177e4 | 240 | { |
d8d1f30b | 241 | struct dst_entry *dst = &rt->dst; |
92d86829 | 242 | bool rc = true; |
1da177e4 LT |
243 | |
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)) | |
e905a9ed | 253 | goto out; |
1da177e4 LT |
254 | |
255 | /* Limit if icmp type is enabled in ratemask. */ | |
92d86829 | 256 | if ((1 << type) & net->ipv4.sysctl_icmp_ratemask) { |
1d861aa4 | 257 | struct inet_peer *peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, 1); |
fbfe95a4 | 258 | rc = inet_peer_xrlim_allow(peer, |
92d86829 | 259 | net->ipv4.sysctl_icmp_ratelimit); |
e1a67642 NC |
260 | if (peer) |
261 | inet_putpeer(peer); | |
92d86829 | 262 | } |
1da177e4 LT |
263 | out: |
264 | return rc; | |
265 | } | |
266 | ||
267 | /* | |
268 | * Maintain the counters used in the SNMP statistics for outgoing ICMP | |
269 | */ | |
0388b004 | 270 | void icmp_out_count(struct net *net, unsigned char type) |
1da177e4 | 271 | { |
903fc196 | 272 | ICMPMSGOUT_INC_STATS(net, type); |
75c939bb | 273 | ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS); |
1da177e4 LT |
274 | } |
275 | ||
276 | /* | |
277 | * Checksum each fragment, and on the first include the headers and final | |
278 | * checksum. | |
279 | */ | |
280 | static 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; | |
5f92a738 | 284 | __wsum csum; |
1da177e4 LT |
285 | |
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 | ||
296 | static void icmp_push_reply(struct icmp_bxm *icmp_param, | |
77968b78 | 297 | struct flowi4 *fl4, |
2e77d89b | 298 | struct ipcm_cookie *ipc, struct rtable **rt) |
1da177e4 | 299 | { |
1e3cf683 | 300 | struct sock *sk; |
1da177e4 LT |
301 | struct sk_buff *skb; |
302 | ||
d8d1f30b | 303 | sk = icmp_sk(dev_net((*rt)->dst.dev)); |
f5fca608 | 304 | if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param, |
e905a9ed YH |
305 | icmp_param->data_len+icmp_param->head_len, |
306 | icmp_param->head_len, | |
1f8438a8 ED |
307 | ipc, rt, MSG_DONTWAIT) < 0) { |
308 | ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS); | |
1e3cf683 | 309 | ip_flush_pending_frames(sk); |
1f8438a8 | 310 | } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) { |
88c7664f | 311 | struct icmphdr *icmph = icmp_hdr(skb); |
d3bc23e7 | 312 | __wsum csum = 0; |
1da177e4 LT |
313 | struct sk_buff *skb1; |
314 | ||
1e3cf683 | 315 | skb_queue_walk(&sk->sk_write_queue, skb1) { |
1da177e4 LT |
316 | 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; | |
77968b78 | 323 | ip_push_pending_frames(sk, fl4); |
1da177e4 LT |
324 | } |
325 | } | |
326 | ||
327 | /* | |
328 | * Driving logic for building and sending ICMP messages. | |
329 | */ | |
330 | ||
331 | static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb) | |
332 | { | |
1da177e4 | 333 | struct ipcm_cookie ipc; |
511c3f92 | 334 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 335 | struct net *net = dev_net(rt->dst.dev); |
77968b78 | 336 | struct flowi4 fl4; |
fdc0bde9 DL |
337 | struct sock *sk; |
338 | struct inet_sock *inet; | |
35ebf65e | 339 | __be32 daddr, saddr; |
1da177e4 | 340 | |
f6d8bd05 | 341 | if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb)) |
f00c401b | 342 | return; |
1da177e4 | 343 | |
fdc0bde9 DL |
344 | sk = icmp_xmit_lock(net); |
345 | if (sk == NULL) | |
1da177e4 | 346 | return; |
fdc0bde9 | 347 | inet = inet_sk(sk); |
1da177e4 LT |
348 | |
349 | icmp_param->data.icmph.checksum = 0; | |
1da177e4 | 350 | |
eddc9ec5 | 351 | inet->tos = ip_hdr(skb)->tos; |
9f6abb5f | 352 | daddr = ipc.addr = ip_hdr(skb)->saddr; |
35ebf65e | 353 | saddr = fib_compute_spec_dst(skb); |
1da177e4 | 354 | ipc.opt = NULL; |
2244d07b | 355 | ipc.tx_flags = 0; |
f6d8bd05 ED |
356 | 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; | |
1da177e4 | 360 | } |
77968b78 DM |
361 | memset(&fl4, 0, sizeof(fl4)); |
362 | fl4.daddr = daddr; | |
35ebf65e | 363 | fl4.saddr = saddr; |
77968b78 DM |
364 | 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; | |
a48eff12 | 370 | if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type, |
1da177e4 | 371 | icmp_param->data.icmph.code)) |
77968b78 | 372 | icmp_push_reply(icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
373 | ip_rt_put(rt); |
374 | out_unlock: | |
405666db | 375 | icmp_xmit_unlock(sk); |
1da177e4 LT |
376 | } |
377 | ||
77968b78 DM |
378 | static struct rtable *icmp_route_lookup(struct net *net, |
379 | struct flowi4 *fl4, | |
380 | struct sk_buff *skb_in, | |
b71d1d42 | 381 | const struct iphdr *iph, |
f6d460cf DM |
382 | __be32 saddr, u8 tos, |
383 | int type, int code, | |
384 | struct icmp_bxm *param) | |
385 | { | |
f6d460cf | 386 | struct rtable *rt, *rt2; |
415b3334 | 387 | struct flowi4 fl4_dec; |
f6d460cf DM |
388 | int err; |
389 | ||
77968b78 DM |
390 | 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); | |
b23dd4fe DM |
400 | if (IS_ERR(rt)) |
401 | return rt; | |
f6d460cf DM |
402 | |
403 | /* No need to clone since we're just using its address. */ | |
404 | rt2 = rt; | |
405 | ||
9d6ec938 | 406 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
77968b78 | 407 | flowi4_to_flowi(fl4), NULL, 0); |
452edd59 | 408 | if (!IS_ERR(rt)) { |
f6d460cf DM |
409 | if (rt != rt2) |
410 | return rt; | |
452edd59 | 411 | } else if (PTR_ERR(rt) == -EPERM) { |
f6d460cf | 412 | rt = NULL; |
452edd59 DM |
413 | } else |
414 | return rt; | |
f6d460cf | 415 | |
415b3334 | 416 | err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET); |
f6d460cf DM |
417 | if (err) |
418 | goto relookup_failed; | |
419 | ||
415b3334 DM |
420 | if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) { |
421 | rt2 = __ip_route_output_key(net, &fl4_dec); | |
b23dd4fe DM |
422 | if (IS_ERR(rt2)) |
423 | err = PTR_ERR(rt2); | |
f6d460cf | 424 | } else { |
9d6ec938 | 425 | struct flowi4 fl4_2 = {}; |
f6d460cf DM |
426 | unsigned long orefdst; |
427 | ||
415b3334 | 428 | fl4_2.daddr = fl4_dec.saddr; |
9d6ec938 | 429 | rt2 = ip_route_output_key(net, &fl4_2); |
b23dd4fe DM |
430 | if (IS_ERR(rt2)) { |
431 | err = PTR_ERR(rt2); | |
f6d460cf | 432 | goto relookup_failed; |
b23dd4fe | 433 | } |
f6d460cf DM |
434 | /* Ugh! */ |
435 | orefdst = skb_in->_skb_refdst; /* save old refdst */ | |
415b3334 | 436 | err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr, |
f6d460cf DM |
437 | 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 | ||
9d6ec938 | 447 | rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst, |
415b3334 | 448 | flowi4_to_flowi(&fl4_dec), NULL, |
9d6ec938 | 449 | XFRM_LOOKUP_ICMP); |
452edd59 | 450 | if (!IS_ERR(rt2)) { |
f6d460cf | 451 | dst_release(&rt->dst); |
415b3334 | 452 | memcpy(fl4, &fl4_dec, sizeof(*fl4)); |
f6d460cf | 453 | rt = rt2; |
452edd59 DM |
454 | } 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; | |
f6d460cf | 461 | } |
f6d460cf DM |
462 | return rt; |
463 | ||
464 | relookup_failed: | |
465 | if (rt) | |
466 | return rt; | |
467 | return ERR_PTR(err); | |
468 | } | |
1da177e4 LT |
469 | |
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 | ||
e4883014 | 481 | void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info) |
1da177e4 LT |
482 | { |
483 | struct iphdr *iph; | |
484 | int room; | |
485 | struct icmp_bxm icmp_param; | |
511c3f92 | 486 | struct rtable *rt = skb_rtable(skb_in); |
1da177e4 | 487 | struct ipcm_cookie ipc; |
77968b78 | 488 | struct flowi4 fl4; |
a61ced5d | 489 | __be32 saddr; |
1da177e4 | 490 | u8 tos; |
dde1bc0e | 491 | struct net *net; |
4a6ad7a1 | 492 | struct sock *sk; |
1da177e4 LT |
493 | |
494 | if (!rt) | |
495 | goto out; | |
d8d1f30b | 496 | net = dev_net(rt->dst.dev); |
1da177e4 LT |
497 | |
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 | */ | |
eddc9ec5 | 503 | iph = ip_hdr(skb_in); |
1da177e4 | 504 | |
27a884dc | 505 | if ((u8 *)iph < skb_in->head || |
f7c0c2ae SH |
506 | (skb_network_header(skb_in) + sizeof(*iph)) > |
507 | skb_tail_pointer(skb_in)) | |
1da177e4 LT |
508 | goto out; |
509 | ||
510 | /* | |
511 | * No replies to physical multicast/broadcast | |
512 | */ | |
513 | if (skb_in->pkt_type != PACKET_HOST) | |
514 | goto out; | |
515 | ||
516 | /* | |
517 | * Now check at the protocol level | |
518 | */ | |
519 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
520 | goto out; | |
521 | ||
522 | /* | |
523 | * Only reply to fragment 0. We byte re-order the constant | |
524 | * mask for efficiency. | |
525 | */ | |
526 | if (iph->frag_off & htons(IP_OFFSET)) | |
527 | goto out; | |
528 | ||
529 | /* | |
530 | * If we send an ICMP error to an ICMP error a mess would result.. | |
531 | */ | |
532 | if (icmp_pointers[type].error) { | |
533 | /* | |
534 | * We are an error, check if we are replying to an | |
535 | * ICMP error | |
536 | */ | |
537 | if (iph->protocol == IPPROTO_ICMP) { | |
538 | u8 _inner_type, *itp; | |
539 | ||
540 | itp = skb_header_pointer(skb_in, | |
d56f90a7 | 541 | skb_network_header(skb_in) + |
1da177e4 LT |
542 | (iph->ihl << 2) + |
543 | offsetof(struct icmphdr, | |
544 | type) - | |
545 | skb_in->data, | |
546 | sizeof(_inner_type), | |
547 | &_inner_type); | |
548 | if (itp == NULL) | |
549 | goto out; | |
550 | ||
551 | /* | |
552 | * Assume any unknown ICMP type is an error. This | |
553 | * isn't specified by the RFC, but think about it.. | |
554 | */ | |
555 | if (*itp > NR_ICMP_TYPES || | |
556 | icmp_pointers[*itp].error) | |
557 | goto out; | |
558 | } | |
559 | } | |
560 | ||
fdc0bde9 DL |
561 | sk = icmp_xmit_lock(net); |
562 | if (sk == NULL) | |
1da177e4 LT |
563 | return; |
564 | ||
565 | /* | |
566 | * Construct source address and options. | |
567 | */ | |
568 | ||
569 | saddr = iph->daddr; | |
1c2fb7f9 | 570 | if (!(rt->rt_flags & RTCF_LOCAL)) { |
6e1d9103 PM |
571 | struct net_device *dev = NULL; |
572 | ||
685c7944 | 573 | rcu_read_lock(); |
c7537967 DM |
574 | if (rt_is_input_route(rt) && |
575 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr) | |
92101b3b | 576 | dev = dev_get_by_index_rcu(net, inet_iif(skb_in)); |
6e1d9103 | 577 | |
685c7944 | 578 | if (dev) |
6e1d9103 | 579 | saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
685c7944 | 580 | else |
1c2fb7f9 | 581 | saddr = 0; |
685c7944 | 582 | rcu_read_unlock(); |
1c2fb7f9 | 583 | } |
1da177e4 LT |
584 | |
585 | tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) | | |
586 | IPTOS_PREC_INTERNETCONTROL) : | |
587 | iph->tos; | |
588 | ||
f6d8bd05 | 589 | if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in)) |
fa60cf7f | 590 | goto out_unlock; |
1da177e4 LT |
591 | |
592 | ||
593 | /* | |
594 | * Prepare data for ICMP header. | |
595 | */ | |
596 | ||
597 | icmp_param.data.icmph.type = type; | |
598 | icmp_param.data.icmph.code = code; | |
599 | icmp_param.data.icmph.un.gateway = info; | |
600 | icmp_param.data.icmph.checksum = 0; | |
601 | icmp_param.skb = skb_in; | |
d56f90a7 | 602 | icmp_param.offset = skb_network_offset(skb_in); |
405666db | 603 | inet_sk(sk)->tos = tos; |
1da177e4 | 604 | ipc.addr = iph->saddr; |
f6d8bd05 | 605 | ipc.opt = &icmp_param.replyopts.opt; |
2244d07b | 606 | ipc.tx_flags = 0; |
1da177e4 | 607 | |
77968b78 | 608 | rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, |
f6d460cf DM |
609 | type, code, &icmp_param); |
610 | if (IS_ERR(rt)) | |
611 | goto out_unlock; | |
1da177e4 | 612 | |
a48eff12 | 613 | if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
1da177e4 LT |
614 | goto ende; |
615 | ||
616 | /* RFC says return as much as we can without exceeding 576 bytes. */ | |
617 | ||
d8d1f30b | 618 | room = dst_mtu(&rt->dst); |
1da177e4 LT |
619 | if (room > 576) |
620 | room = 576; | |
f6d8bd05 | 621 | room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen; |
1da177e4 LT |
622 | room -= sizeof(struct icmphdr); |
623 | ||
624 | icmp_param.data_len = skb_in->len - icmp_param.offset; | |
625 | if (icmp_param.data_len > room) | |
626 | icmp_param.data_len = room; | |
627 | icmp_param.head_len = sizeof(struct icmphdr); | |
628 | ||
77968b78 | 629 | icmp_push_reply(&icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
630 | ende: |
631 | ip_rt_put(rt); | |
632 | out_unlock: | |
405666db | 633 | icmp_xmit_unlock(sk); |
1da177e4 LT |
634 | out:; |
635 | } | |
4bc2f18b | 636 | EXPORT_SYMBOL(icmp_send); |
1da177e4 LT |
637 | |
638 | ||
1de9243b DM |
639 | static void icmp_socket_deliver(struct sk_buff *skb, u32 info) |
640 | { | |
641 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
642 | const struct net_protocol *ipprot; | |
643 | int protocol = iph->protocol; | |
644 | ||
f0a70e90 DM |
645 | /* Checkin full IP header plus 8 bytes of protocol to |
646 | * avoid additional coding at protocol handlers. | |
647 | */ | |
648 | if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) | |
649 | return; | |
650 | ||
1de9243b DM |
651 | raw_icmp_error(skb, protocol, info); |
652 | ||
653 | rcu_read_lock(); | |
654 | ipprot = rcu_dereference(inet_protos[protocol]); | |
655 | if (ipprot && ipprot->err_handler) | |
656 | ipprot->err_handler(skb, info); | |
657 | rcu_read_unlock(); | |
658 | } | |
659 | ||
1da177e4 LT |
660 | /* |
661 | * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH. | |
662 | */ | |
663 | ||
664 | static void icmp_unreach(struct sk_buff *skb) | |
665 | { | |
b71d1d42 | 666 | const struct iphdr *iph; |
1da177e4 | 667 | struct icmphdr *icmph; |
dde1bc0e | 668 | struct net *net; |
f9242b6b | 669 | u32 info = 0; |
dde1bc0e | 670 | |
adf30907 | 671 | net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
672 | |
673 | /* | |
674 | * Incomplete header ? | |
675 | * Only checks for the IP header, there should be an | |
676 | * additional check for longer headers in upper levels. | |
677 | */ | |
678 | ||
679 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
680 | goto out_err; | |
681 | ||
88c7664f | 682 | icmph = icmp_hdr(skb); |
b71d1d42 | 683 | iph = (const struct iphdr *)skb->data; |
1da177e4 LT |
684 | |
685 | if (iph->ihl < 5) /* Mangled header, drop. */ | |
686 | goto out_err; | |
687 | ||
688 | if (icmph->type == ICMP_DEST_UNREACH) { | |
689 | switch (icmph->code & 15) { | |
690 | case ICMP_NET_UNREACH: | |
691 | case ICMP_HOST_UNREACH: | |
692 | case ICMP_PROT_UNREACH: | |
693 | case ICMP_PORT_UNREACH: | |
694 | break; | |
695 | case ICMP_FRAG_NEEDED: | |
696 | if (ipv4_config.no_pmtu_disc) { | |
afd46503 | 697 | LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"), |
673d57e7 | 698 | &iph->daddr); |
1da177e4 | 699 | } else { |
46517008 | 700 | info = ntohs(icmph->un.frag.mtu); |
1da177e4 LT |
701 | if (!info) |
702 | goto out; | |
703 | } | |
704 | break; | |
705 | case ICMP_SR_FAILED: | |
afd46503 | 706 | LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"), |
673d57e7 | 707 | &iph->daddr); |
1da177e4 LT |
708 | break; |
709 | default: | |
710 | break; | |
711 | } | |
712 | if (icmph->code > NR_ICMP_UNREACH) | |
713 | goto out; | |
714 | } else if (icmph->type == ICMP_PARAMETERPROB) | |
715 | info = ntohl(icmph->un.gateway) >> 24; | |
716 | ||
717 | /* | |
718 | * Throw it at our lower layers | |
719 | * | |
720 | * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed | |
721 | * header. | |
722 | * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the | |
723 | * transport layer. | |
724 | * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to | |
725 | * transport layer. | |
726 | */ | |
727 | ||
728 | /* | |
25985edc | 729 | * Check the other end isn't violating RFC 1122. Some routers send |
1da177e4 LT |
730 | * bogus responses to broadcast frames. If you see this message |
731 | * first check your netmask matches at both ends, if it does then | |
732 | * get the other vendor to fix their kit. | |
733 | */ | |
734 | ||
b34a95ee | 735 | if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses && |
dde1bc0e | 736 | inet_addr_type(net, iph->daddr) == RTN_BROADCAST) { |
e87cc472 JP |
737 | net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n", |
738 | &ip_hdr(skb)->saddr, | |
739 | icmph->type, icmph->code, | |
740 | &iph->daddr, skb->dev->name); | |
1da177e4 LT |
741 | goto out; |
742 | } | |
743 | ||
1de9243b | 744 | icmp_socket_deliver(skb, info); |
1da177e4 LT |
745 | |
746 | out: | |
747 | return; | |
748 | out_err: | |
dcfc23ca | 749 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
750 | goto out; |
751 | } | |
752 | ||
753 | ||
754 | /* | |
755 | * Handle ICMP_REDIRECT. | |
756 | */ | |
757 | ||
758 | static void icmp_redirect(struct sk_buff *skb) | |
759 | { | |
94206125 DM |
760 | if (skb->len < sizeof(struct iphdr)) { |
761 | ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS); | |
762 | return; | |
763 | } | |
1da177e4 | 764 | |
1da177e4 | 765 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
94206125 | 766 | return; |
c319b4d7 | 767 | |
d3351b75 | 768 | icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway); |
1da177e4 LT |
769 | } |
770 | ||
771 | /* | |
772 | * Handle ICMP_ECHO ("ping") requests. | |
773 | * | |
774 | * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo | |
775 | * requests. | |
776 | * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be | |
777 | * included in the reply. | |
778 | * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring | |
779 | * echo requests, MUST have default=NOT. | |
780 | * See also WRT handling of options once they are done and working. | |
781 | */ | |
782 | ||
783 | static void icmp_echo(struct sk_buff *skb) | |
784 | { | |
b34a95ee PE |
785 | struct net *net; |
786 | ||
adf30907 | 787 | net = dev_net(skb_dst(skb)->dev); |
b34a95ee | 788 | if (!net->ipv4.sysctl_icmp_echo_ignore_all) { |
1da177e4 LT |
789 | struct icmp_bxm icmp_param; |
790 | ||
88c7664f | 791 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
792 | icmp_param.data.icmph.type = ICMP_ECHOREPLY; |
793 | icmp_param.skb = skb; | |
794 | icmp_param.offset = 0; | |
795 | icmp_param.data_len = skb->len; | |
796 | icmp_param.head_len = sizeof(struct icmphdr); | |
797 | icmp_reply(&icmp_param, skb); | |
798 | } | |
799 | } | |
800 | ||
801 | /* | |
802 | * Handle ICMP Timestamp requests. | |
803 | * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests. | |
804 | * SHOULD be in the kernel for minimum random latency. | |
805 | * MUST be accurate to a few minutes. | |
806 | * MUST be updated at least at 15Hz. | |
807 | */ | |
808 | static void icmp_timestamp(struct sk_buff *skb) | |
809 | { | |
f25c3d61 | 810 | struct timespec tv; |
1da177e4 LT |
811 | struct icmp_bxm icmp_param; |
812 | /* | |
813 | * Too short. | |
814 | */ | |
815 | if (skb->len < 4) | |
816 | goto out_err; | |
817 | ||
818 | /* | |
819 | * Fill in the current time as ms since midnight UT: | |
820 | */ | |
f25c3d61 YH |
821 | getnstimeofday(&tv); |
822 | icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + | |
823 | tv.tv_nsec / NSEC_PER_MSEC); | |
1da177e4 LT |
824 | icmp_param.data.times[2] = icmp_param.data.times[1]; |
825 | if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4)) | |
826 | BUG(); | |
88c7664f | 827 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
828 | icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY; |
829 | icmp_param.data.icmph.code = 0; | |
830 | icmp_param.skb = skb; | |
831 | icmp_param.offset = 0; | |
832 | icmp_param.data_len = 0; | |
833 | icmp_param.head_len = sizeof(struct icmphdr) + 12; | |
834 | icmp_reply(&icmp_param, skb); | |
835 | out: | |
836 | return; | |
837 | out_err: | |
adf30907 | 838 | ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS); |
1da177e4 LT |
839 | goto out; |
840 | } | |
841 | ||
1da177e4 LT |
842 | static void icmp_discard(struct sk_buff *skb) |
843 | { | |
844 | } | |
845 | ||
846 | /* | |
847 | * Deal with incoming ICMP packets. | |
848 | */ | |
849 | int icmp_rcv(struct sk_buff *skb) | |
850 | { | |
851 | struct icmphdr *icmph; | |
511c3f92 | 852 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 853 | struct net *net = dev_net(rt->dst.dev); |
1da177e4 | 854 | |
aebcf82c | 855 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { |
def8b4fa | 856 | struct sec_path *sp = skb_sec_path(skb); |
8b7817f3 HX |
857 | int nh; |
858 | ||
def8b4fa | 859 | if (!(sp && sp->xvec[sp->len - 1]->props.flags & |
aebcf82c HX |
860 | XFRM_STATE_ICMP)) |
861 | goto drop; | |
862 | ||
8b7817f3 HX |
863 | if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr))) |
864 | goto drop; | |
865 | ||
866 | nh = skb_network_offset(skb); | |
867 | skb_set_network_header(skb, sizeof(*icmph)); | |
868 | ||
869 | if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) | |
870 | goto drop; | |
871 | ||
872 | skb_set_network_header(skb, nh); | |
873 | } | |
874 | ||
dcfc23ca | 875 | ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS); |
1da177e4 LT |
876 | |
877 | switch (skb->ip_summed) { | |
84fa7933 | 878 | case CHECKSUM_COMPLETE: |
d3bc23e7 | 879 | if (!csum_fold(skb->csum)) |
1da177e4 | 880 | break; |
fb286bb2 | 881 | /* fall through */ |
1da177e4 | 882 | case CHECKSUM_NONE: |
fb286bb2 HX |
883 | skb->csum = 0; |
884 | if (__skb_checksum_complete(skb)) | |
6a5dc9e5 | 885 | goto csum_error; |
1da177e4 LT |
886 | } |
887 | ||
8cf22943 HX |
888 | if (!pskb_pull(skb, sizeof(*icmph))) |
889 | goto error; | |
1da177e4 | 890 | |
88c7664f | 891 | icmph = icmp_hdr(skb); |
1da177e4 | 892 | |
f66ac03d | 893 | ICMPMSGIN_INC_STATS_BH(net, icmph->type); |
1da177e4 LT |
894 | /* |
895 | * 18 is the highest 'known' ICMP type. Anything else is a mystery | |
896 | * | |
897 | * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently | |
898 | * discarded. | |
899 | */ | |
900 | if (icmph->type > NR_ICMP_TYPES) | |
901 | goto error; | |
902 | ||
903 | ||
904 | /* | |
905 | * Parse the ICMP message | |
906 | */ | |
907 | ||
e905a9ed | 908 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
1da177e4 LT |
909 | /* |
910 | * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be | |
911 | * silently ignored (we let user decide with a sysctl). | |
912 | * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently | |
913 | * discarded if to broadcast/multicast. | |
914 | */ | |
4c866aa7 AK |
915 | if ((icmph->type == ICMP_ECHO || |
916 | icmph->type == ICMP_TIMESTAMP) && | |
b34a95ee | 917 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts) { |
1da177e4 LT |
918 | goto error; |
919 | } | |
920 | if (icmph->type != ICMP_ECHO && | |
921 | icmph->type != ICMP_TIMESTAMP && | |
922 | icmph->type != ICMP_ADDRESS && | |
923 | icmph->type != ICMP_ADDRESSREPLY) { | |
924 | goto error; | |
e905a9ed | 925 | } |
1da177e4 LT |
926 | } |
927 | ||
1da177e4 LT |
928 | icmp_pointers[icmph->type].handler(skb); |
929 | ||
930 | drop: | |
931 | kfree_skb(skb); | |
932 | return 0; | |
6a5dc9e5 ED |
933 | csum_error: |
934 | ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS); | |
1da177e4 | 935 | error: |
dcfc23ca | 936 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
937 | goto drop; |
938 | } | |
939 | ||
5b052042 LW |
940 | void icmp_err(struct sk_buff *skb, u32 info) |
941 | { | |
942 | struct iphdr *iph = (struct iphdr *)skb->data; | |
6d0bfe22 LC |
943 | int offset = iph->ihl<<2; |
944 | struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset); | |
5b052042 LW |
945 | int type = icmp_hdr(skb)->type; |
946 | int code = icmp_hdr(skb)->code; | |
947 | struct net *net = dev_net(skb->dev); | |
948 | ||
949 | /* | |
950 | * Use ping_err to handle all icmp errors except those | |
951 | * triggered by ICMP_ECHOREPLY which sent from kernel. | |
952 | */ | |
953 | if (icmph->type != ICMP_ECHOREPLY) { | |
6d0bfe22 | 954 | ping_err(skb, offset, info); |
5b052042 LW |
955 | return; |
956 | } | |
957 | ||
958 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
959 | ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0); | |
960 | else if (type == ICMP_REDIRECT) | |
961 | ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0); | |
962 | } | |
963 | ||
1da177e4 LT |
964 | /* |
965 | * This table is the definition of how we handle ICMP. | |
966 | */ | |
9b5b5cff | 967 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = { |
1da177e4 | 968 | [ICMP_ECHOREPLY] = { |
c319b4d7 | 969 | .handler = ping_rcv, |
1da177e4 LT |
970 | }, |
971 | [1] = { | |
1da177e4 LT |
972 | .handler = icmp_discard, |
973 | .error = 1, | |
974 | }, | |
975 | [2] = { | |
1da177e4 LT |
976 | .handler = icmp_discard, |
977 | .error = 1, | |
978 | }, | |
979 | [ICMP_DEST_UNREACH] = { | |
1da177e4 LT |
980 | .handler = icmp_unreach, |
981 | .error = 1, | |
982 | }, | |
983 | [ICMP_SOURCE_QUENCH] = { | |
1da177e4 LT |
984 | .handler = icmp_unreach, |
985 | .error = 1, | |
986 | }, | |
987 | [ICMP_REDIRECT] = { | |
1da177e4 LT |
988 | .handler = icmp_redirect, |
989 | .error = 1, | |
990 | }, | |
991 | [6] = { | |
1da177e4 LT |
992 | .handler = icmp_discard, |
993 | .error = 1, | |
994 | }, | |
995 | [7] = { | |
1da177e4 LT |
996 | .handler = icmp_discard, |
997 | .error = 1, | |
998 | }, | |
999 | [ICMP_ECHO] = { | |
1da177e4 LT |
1000 | .handler = icmp_echo, |
1001 | }, | |
1002 | [9] = { | |
1da177e4 LT |
1003 | .handler = icmp_discard, |
1004 | .error = 1, | |
1005 | }, | |
1006 | [10] = { | |
1da177e4 LT |
1007 | .handler = icmp_discard, |
1008 | .error = 1, | |
1009 | }, | |
1010 | [ICMP_TIME_EXCEEDED] = { | |
1da177e4 LT |
1011 | .handler = icmp_unreach, |
1012 | .error = 1, | |
1013 | }, | |
1014 | [ICMP_PARAMETERPROB] = { | |
1da177e4 LT |
1015 | .handler = icmp_unreach, |
1016 | .error = 1, | |
1017 | }, | |
1018 | [ICMP_TIMESTAMP] = { | |
1da177e4 LT |
1019 | .handler = icmp_timestamp, |
1020 | }, | |
1021 | [ICMP_TIMESTAMPREPLY] = { | |
1da177e4 LT |
1022 | .handler = icmp_discard, |
1023 | }, | |
1024 | [ICMP_INFO_REQUEST] = { | |
1da177e4 LT |
1025 | .handler = icmp_discard, |
1026 | }, | |
e905a9ed | 1027 | [ICMP_INFO_REPLY] = { |
1da177e4 LT |
1028 | .handler = icmp_discard, |
1029 | }, | |
1030 | [ICMP_ADDRESS] = { | |
838942a5 | 1031 | .handler = icmp_discard, |
1da177e4 LT |
1032 | }, |
1033 | [ICMP_ADDRESSREPLY] = { | |
838942a5 | 1034 | .handler = icmp_discard, |
1da177e4 LT |
1035 | }, |
1036 | }; | |
1037 | ||
4a6ad7a1 | 1038 | static void __net_exit icmp_sk_exit(struct net *net) |
1da177e4 | 1039 | { |
1da177e4 LT |
1040 | int i; |
1041 | ||
5c8cafd6 | 1042 | for_each_possible_cpu(i) |
c1e9894d | 1043 | inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]); |
4a6ad7a1 DL |
1044 | kfree(net->ipv4.icmp_sk); |
1045 | net->ipv4.icmp_sk = NULL; | |
a5710d65 | 1046 | } |
1da177e4 | 1047 | |
263173af | 1048 | static int __net_init icmp_sk_init(struct net *net) |
a5710d65 | 1049 | { |
a5710d65 DL |
1050 | int i, err; |
1051 | ||
4a6ad7a1 DL |
1052 | net->ipv4.icmp_sk = |
1053 | kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL); | |
1054 | if (net->ipv4.icmp_sk == NULL) | |
79c91159 DL |
1055 | return -ENOMEM; |
1056 | ||
a5710d65 | 1057 | for_each_possible_cpu(i) { |
1e3cf683 | 1058 | struct sock *sk; |
1da177e4 | 1059 | |
c1e9894d DL |
1060 | err = inet_ctl_sock_create(&sk, PF_INET, |
1061 | SOCK_RAW, IPPROTO_ICMP, net); | |
1da177e4 | 1062 | if (err < 0) |
a5710d65 | 1063 | goto fail; |
1da177e4 | 1064 | |
c1e9894d | 1065 | net->ipv4.icmp_sk[i] = sk; |
1da177e4 LT |
1066 | |
1067 | /* Enough space for 2 64K ICMP packets, including | |
87fb4b7b | 1068 | * sk_buff/skb_shared_info struct overhead. |
1da177e4 | 1069 | */ |
87fb4b7b | 1070 | sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); |
1da177e4 | 1071 | |
b3a5b6cc ED |
1072 | /* |
1073 | * Speedup sock_wfree() | |
1074 | */ | |
1075 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
c1e9894d | 1076 | inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT; |
1da177e4 | 1077 | } |
a24022e1 PE |
1078 | |
1079 | /* Control parameters for ECHO replies. */ | |
1080 | net->ipv4.sysctl_icmp_echo_ignore_all = 0; | |
1081 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1; | |
1082 | ||
1083 | /* Control parameter - ignore bogus broadcast responses? */ | |
1084 | net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1; | |
1085 | ||
1086 | /* | |
1087 | * Configurable global rate limit. | |
1088 | * | |
1089 | * ratelimit defines tokens/packet consumed for dst->rate_token | |
1090 | * bucket ratemask defines which icmp types are ratelimited by | |
1091 | * setting it's bit position. | |
1092 | * | |
1093 | * default: | |
1094 | * dest unreachable (3), source quench (4), | |
1095 | * time exceeded (11), parameter problem (12) | |
1096 | */ | |
1097 | ||
1098 | net->ipv4.sysctl_icmp_ratelimit = 1 * HZ; | |
1099 | net->ipv4.sysctl_icmp_ratemask = 0x1818; | |
1100 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0; | |
1101 | ||
a5710d65 DL |
1102 | return 0; |
1103 | ||
1104 | fail: | |
1d1c8d13 | 1105 | for_each_possible_cpu(i) |
c1e9894d | 1106 | inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]); |
1d1c8d13 | 1107 | kfree(net->ipv4.icmp_sk); |
a5710d65 | 1108 | return err; |
1da177e4 LT |
1109 | } |
1110 | ||
4a6ad7a1 DL |
1111 | static struct pernet_operations __net_initdata icmp_sk_ops = { |
1112 | .init = icmp_sk_init, | |
1113 | .exit = icmp_sk_exit, | |
1114 | }; | |
1115 | ||
1116 | int __init icmp_init(void) | |
1117 | { | |
959d2726 | 1118 | return register_pernet_subsys(&icmp_sk_ops); |
4a6ad7a1 | 1119 | } |