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> |
1da177e4 LT |
98 | |
99 | /* | |
100 | * Build xmit assembly blocks | |
101 | */ | |
102 | ||
103 | struct icmp_bxm { | |
104 | struct sk_buff *skb; | |
105 | int offset; | |
106 | int data_len; | |
107 | ||
108 | struct { | |
109 | struct icmphdr icmph; | |
b03d73e3 | 110 | __be32 times[3]; |
1da177e4 LT |
111 | } data; |
112 | int head_len; | |
f6d8bd05 | 113 | struct ip_options_data replyopts; |
1da177e4 LT |
114 | }; |
115 | ||
1da177e4 LT |
116 | /* An array of errno for error messages from dest unreach. */ |
117 | /* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */ | |
118 | ||
e754834e | 119 | const struct icmp_err icmp_err_convert[] = { |
1da177e4 LT |
120 | { |
121 | .errno = ENETUNREACH, /* ICMP_NET_UNREACH */ | |
122 | .fatal = 0, | |
123 | }, | |
124 | { | |
125 | .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */ | |
126 | .fatal = 0, | |
127 | }, | |
128 | { | |
129 | .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */, | |
130 | .fatal = 1, | |
131 | }, | |
132 | { | |
133 | .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */ | |
134 | .fatal = 1, | |
135 | }, | |
136 | { | |
137 | .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */ | |
138 | .fatal = 0, | |
139 | }, | |
140 | { | |
141 | .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */ | |
142 | .fatal = 0, | |
143 | }, | |
144 | { | |
145 | .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */ | |
146 | .fatal = 1, | |
147 | }, | |
148 | { | |
149 | .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */ | |
150 | .fatal = 1, | |
151 | }, | |
152 | { | |
153 | .errno = ENONET, /* ICMP_HOST_ISOLATED */ | |
154 | .fatal = 1, | |
155 | }, | |
156 | { | |
157 | .errno = ENETUNREACH, /* ICMP_NET_ANO */ | |
158 | .fatal = 1, | |
159 | }, | |
160 | { | |
161 | .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */ | |
162 | .fatal = 1, | |
163 | }, | |
164 | { | |
165 | .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */ | |
166 | .fatal = 0, | |
167 | }, | |
168 | { | |
169 | .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */ | |
170 | .fatal = 0, | |
171 | }, | |
172 | { | |
173 | .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */ | |
174 | .fatal = 1, | |
175 | }, | |
176 | { | |
177 | .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */ | |
178 | .fatal = 1, | |
179 | }, | |
180 | { | |
181 | .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */ | |
182 | .fatal = 1, | |
183 | }, | |
184 | }; | |
4bc2f18b | 185 | EXPORT_SYMBOL(icmp_err_convert); |
1da177e4 | 186 | |
1da177e4 LT |
187 | /* |
188 | * ICMP control array. This specifies what to do with each ICMP. | |
189 | */ | |
190 | ||
191 | struct icmp_control { | |
1da177e4 LT |
192 | void (*handler)(struct sk_buff *skb); |
193 | short error; /* This ICMP is classed as an error message */ | |
194 | }; | |
195 | ||
9b5b5cff | 196 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1]; |
1da177e4 LT |
197 | |
198 | /* | |
199 | * The ICMP socket(s). This is the most convenient way to flow control | |
200 | * our ICMP output as well as maintain a clean interface throughout | |
201 | * all layers. All Socketless IP sends will soon be gone. | |
202 | * | |
203 | * On SMP we have one ICMP socket per-cpu. | |
204 | */ | |
4a6ad7a1 DL |
205 | static struct sock *icmp_sk(struct net *net) |
206 | { | |
207 | return net->ipv4.icmp_sk[smp_processor_id()]; | |
208 | } | |
1da177e4 | 209 | |
fdc0bde9 | 210 | static inline struct sock *icmp_xmit_lock(struct net *net) |
1da177e4 | 211 | { |
fdc0bde9 DL |
212 | struct sock *sk; |
213 | ||
1da177e4 LT |
214 | local_bh_disable(); |
215 | ||
fdc0bde9 DL |
216 | sk = icmp_sk(net); |
217 | ||
405666db | 218 | if (unlikely(!spin_trylock(&sk->sk_lock.slock))) { |
1da177e4 LT |
219 | /* This can happen if the output path signals a |
220 | * dst_link_failure() for an outgoing ICMP packet. | |
221 | */ | |
222 | local_bh_enable(); | |
fdc0bde9 | 223 | return NULL; |
1da177e4 | 224 | } |
fdc0bde9 | 225 | return sk; |
1da177e4 LT |
226 | } |
227 | ||
405666db | 228 | static inline void icmp_xmit_unlock(struct sock *sk) |
1da177e4 | 229 | { |
405666db | 230 | spin_unlock_bh(&sk->sk_lock.slock); |
1da177e4 LT |
231 | } |
232 | ||
233 | /* | |
234 | * Send an ICMP frame. | |
235 | */ | |
236 | ||
92d86829 | 237 | static inline bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt, |
a48eff12 | 238 | struct flowi4 *fl4, int type, int code) |
1da177e4 | 239 | { |
d8d1f30b | 240 | struct dst_entry *dst = &rt->dst; |
92d86829 | 241 | bool rc = true; |
1da177e4 LT |
242 | |
243 | if (type > NR_ICMP_TYPES) | |
244 | goto out; | |
245 | ||
246 | /* Don't limit PMTU discovery. */ | |
247 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
248 | goto out; | |
249 | ||
250 | /* No rate limit on loopback */ | |
251 | if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) | |
e905a9ed | 252 | goto out; |
1da177e4 LT |
253 | |
254 | /* Limit if icmp type is enabled in ratemask. */ | |
92d86829 | 255 | if ((1 << type) & net->ipv4.sysctl_icmp_ratemask) { |
fbfe95a4 DM |
256 | struct inet_peer *peer = rt_get_peer_create(rt, fl4->daddr); |
257 | rc = inet_peer_xrlim_allow(peer, | |
92d86829 DM |
258 | net->ipv4.sysctl_icmp_ratelimit); |
259 | } | |
1da177e4 LT |
260 | out: |
261 | return rc; | |
262 | } | |
263 | ||
264 | /* | |
265 | * Maintain the counters used in the SNMP statistics for outgoing ICMP | |
266 | */ | |
0388b004 | 267 | void icmp_out_count(struct net *net, unsigned char type) |
1da177e4 | 268 | { |
903fc196 | 269 | ICMPMSGOUT_INC_STATS(net, type); |
75c939bb | 270 | ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS); |
1da177e4 LT |
271 | } |
272 | ||
273 | /* | |
274 | * Checksum each fragment, and on the first include the headers and final | |
275 | * checksum. | |
276 | */ | |
277 | static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd, | |
278 | struct sk_buff *skb) | |
279 | { | |
280 | struct icmp_bxm *icmp_param = (struct icmp_bxm *)from; | |
5f92a738 | 281 | __wsum csum; |
1da177e4 LT |
282 | |
283 | csum = skb_copy_and_csum_bits(icmp_param->skb, | |
284 | icmp_param->offset + offset, | |
285 | to, len, 0); | |
286 | ||
287 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
288 | if (icmp_pointers[icmp_param->data.icmph.type].error) | |
289 | nf_ct_attach(skb, icmp_param->skb); | |
290 | return 0; | |
291 | } | |
292 | ||
293 | static void icmp_push_reply(struct icmp_bxm *icmp_param, | |
77968b78 | 294 | struct flowi4 *fl4, |
2e77d89b | 295 | struct ipcm_cookie *ipc, struct rtable **rt) |
1da177e4 | 296 | { |
1e3cf683 | 297 | struct sock *sk; |
1da177e4 LT |
298 | struct sk_buff *skb; |
299 | ||
d8d1f30b | 300 | sk = icmp_sk(dev_net((*rt)->dst.dev)); |
f5fca608 | 301 | if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param, |
e905a9ed YH |
302 | icmp_param->data_len+icmp_param->head_len, |
303 | icmp_param->head_len, | |
1f8438a8 ED |
304 | ipc, rt, MSG_DONTWAIT) < 0) { |
305 | ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS); | |
1e3cf683 | 306 | ip_flush_pending_frames(sk); |
1f8438a8 | 307 | } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) { |
88c7664f | 308 | struct icmphdr *icmph = icmp_hdr(skb); |
d3bc23e7 | 309 | __wsum csum = 0; |
1da177e4 LT |
310 | struct sk_buff *skb1; |
311 | ||
1e3cf683 | 312 | skb_queue_walk(&sk->sk_write_queue, skb1) { |
1da177e4 LT |
313 | csum = csum_add(csum, skb1->csum); |
314 | } | |
315 | csum = csum_partial_copy_nocheck((void *)&icmp_param->data, | |
316 | (char *)icmph, | |
317 | icmp_param->head_len, csum); | |
318 | icmph->checksum = csum_fold(csum); | |
319 | skb->ip_summed = CHECKSUM_NONE; | |
77968b78 | 320 | ip_push_pending_frames(sk, fl4); |
1da177e4 LT |
321 | } |
322 | } | |
323 | ||
324 | /* | |
325 | * Driving logic for building and sending ICMP messages. | |
326 | */ | |
327 | ||
328 | static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb) | |
329 | { | |
1da177e4 | 330 | struct ipcm_cookie ipc; |
511c3f92 | 331 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 332 | struct net *net = dev_net(rt->dst.dev); |
77968b78 | 333 | struct flowi4 fl4; |
fdc0bde9 DL |
334 | struct sock *sk; |
335 | struct inet_sock *inet; | |
3ca3c68e | 336 | __be32 daddr; |
1da177e4 | 337 | |
f6d8bd05 | 338 | if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb)) |
f00c401b | 339 | return; |
1da177e4 | 340 | |
fdc0bde9 DL |
341 | sk = icmp_xmit_lock(net); |
342 | if (sk == NULL) | |
1da177e4 | 343 | return; |
fdc0bde9 | 344 | inet = inet_sk(sk); |
1da177e4 LT |
345 | |
346 | icmp_param->data.icmph.checksum = 0; | |
1da177e4 | 347 | |
eddc9ec5 | 348 | inet->tos = ip_hdr(skb)->tos; |
9f6abb5f | 349 | daddr = ipc.addr = ip_hdr(skb)->saddr; |
1da177e4 | 350 | ipc.opt = NULL; |
2244d07b | 351 | ipc.tx_flags = 0; |
f6d8bd05 ED |
352 | if (icmp_param->replyopts.opt.opt.optlen) { |
353 | ipc.opt = &icmp_param->replyopts.opt; | |
354 | if (ipc.opt->opt.srr) | |
355 | daddr = icmp_param->replyopts.opt.opt.faddr; | |
1da177e4 | 356 | } |
77968b78 DM |
357 | memset(&fl4, 0, sizeof(fl4)); |
358 | fl4.daddr = daddr; | |
359 | fl4.saddr = rt->rt_spec_dst; | |
360 | fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos); | |
361 | fl4.flowi4_proto = IPPROTO_ICMP; | |
362 | security_skb_classify_flow(skb, flowi4_to_flowi(&fl4)); | |
363 | rt = ip_route_output_key(net, &fl4); | |
364 | if (IS_ERR(rt)) | |
365 | goto out_unlock; | |
a48eff12 | 366 | if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type, |
1da177e4 | 367 | icmp_param->data.icmph.code)) |
77968b78 | 368 | icmp_push_reply(icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
369 | ip_rt_put(rt); |
370 | out_unlock: | |
405666db | 371 | icmp_xmit_unlock(sk); |
1da177e4 LT |
372 | } |
373 | ||
77968b78 DM |
374 | static struct rtable *icmp_route_lookup(struct net *net, |
375 | struct flowi4 *fl4, | |
376 | struct sk_buff *skb_in, | |
b71d1d42 | 377 | const struct iphdr *iph, |
f6d460cf DM |
378 | __be32 saddr, u8 tos, |
379 | int type, int code, | |
380 | struct icmp_bxm *param) | |
381 | { | |
f6d460cf | 382 | struct rtable *rt, *rt2; |
415b3334 | 383 | struct flowi4 fl4_dec; |
f6d460cf DM |
384 | int err; |
385 | ||
77968b78 DM |
386 | memset(fl4, 0, sizeof(*fl4)); |
387 | fl4->daddr = (param->replyopts.opt.opt.srr ? | |
388 | param->replyopts.opt.opt.faddr : iph->saddr); | |
389 | fl4->saddr = saddr; | |
390 | fl4->flowi4_tos = RT_TOS(tos); | |
391 | fl4->flowi4_proto = IPPROTO_ICMP; | |
392 | fl4->fl4_icmp_type = type; | |
393 | fl4->fl4_icmp_code = code; | |
394 | security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4)); | |
395 | rt = __ip_route_output_key(net, fl4); | |
b23dd4fe DM |
396 | if (IS_ERR(rt)) |
397 | return rt; | |
f6d460cf DM |
398 | |
399 | /* No need to clone since we're just using its address. */ | |
400 | rt2 = rt; | |
401 | ||
9d6ec938 | 402 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
77968b78 | 403 | flowi4_to_flowi(fl4), NULL, 0); |
452edd59 | 404 | if (!IS_ERR(rt)) { |
f6d460cf DM |
405 | if (rt != rt2) |
406 | return rt; | |
452edd59 | 407 | } else if (PTR_ERR(rt) == -EPERM) { |
f6d460cf | 408 | rt = NULL; |
452edd59 DM |
409 | } else |
410 | return rt; | |
f6d460cf | 411 | |
415b3334 | 412 | err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET); |
f6d460cf DM |
413 | if (err) |
414 | goto relookup_failed; | |
415 | ||
415b3334 DM |
416 | if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) { |
417 | rt2 = __ip_route_output_key(net, &fl4_dec); | |
b23dd4fe DM |
418 | if (IS_ERR(rt2)) |
419 | err = PTR_ERR(rt2); | |
f6d460cf | 420 | } else { |
9d6ec938 | 421 | struct flowi4 fl4_2 = {}; |
f6d460cf DM |
422 | unsigned long orefdst; |
423 | ||
415b3334 | 424 | fl4_2.daddr = fl4_dec.saddr; |
9d6ec938 | 425 | rt2 = ip_route_output_key(net, &fl4_2); |
b23dd4fe DM |
426 | if (IS_ERR(rt2)) { |
427 | err = PTR_ERR(rt2); | |
f6d460cf | 428 | goto relookup_failed; |
b23dd4fe | 429 | } |
f6d460cf DM |
430 | /* Ugh! */ |
431 | orefdst = skb_in->_skb_refdst; /* save old refdst */ | |
415b3334 | 432 | err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr, |
f6d460cf DM |
433 | RT_TOS(tos), rt2->dst.dev); |
434 | ||
435 | dst_release(&rt2->dst); | |
436 | rt2 = skb_rtable(skb_in); | |
437 | skb_in->_skb_refdst = orefdst; /* restore old refdst */ | |
438 | } | |
439 | ||
440 | if (err) | |
441 | goto relookup_failed; | |
442 | ||
9d6ec938 | 443 | rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst, |
415b3334 | 444 | flowi4_to_flowi(&fl4_dec), NULL, |
9d6ec938 | 445 | XFRM_LOOKUP_ICMP); |
452edd59 | 446 | if (!IS_ERR(rt2)) { |
f6d460cf | 447 | dst_release(&rt->dst); |
415b3334 | 448 | memcpy(fl4, &fl4_dec, sizeof(*fl4)); |
f6d460cf | 449 | rt = rt2; |
452edd59 DM |
450 | } else if (PTR_ERR(rt2) == -EPERM) { |
451 | if (rt) | |
452 | dst_release(&rt->dst); | |
453 | return rt2; | |
454 | } else { | |
455 | err = PTR_ERR(rt2); | |
456 | goto relookup_failed; | |
f6d460cf | 457 | } |
f6d460cf DM |
458 | return rt; |
459 | ||
460 | relookup_failed: | |
461 | if (rt) | |
462 | return rt; | |
463 | return ERR_PTR(err); | |
464 | } | |
1da177e4 LT |
465 | |
466 | /* | |
467 | * Send an ICMP message in response to a situation | |
468 | * | |
469 | * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header. | |
470 | * MAY send more (we do). | |
471 | * MUST NOT change this header information. | |
472 | * MUST NOT reply to a multicast/broadcast IP address. | |
473 | * MUST NOT reply to a multicast/broadcast MAC address. | |
474 | * MUST reply to only the first fragment. | |
475 | */ | |
476 | ||
e4883014 | 477 | void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info) |
1da177e4 LT |
478 | { |
479 | struct iphdr *iph; | |
480 | int room; | |
481 | struct icmp_bxm icmp_param; | |
511c3f92 | 482 | struct rtable *rt = skb_rtable(skb_in); |
1da177e4 | 483 | struct ipcm_cookie ipc; |
77968b78 | 484 | struct flowi4 fl4; |
a61ced5d | 485 | __be32 saddr; |
1da177e4 | 486 | u8 tos; |
dde1bc0e | 487 | struct net *net; |
4a6ad7a1 | 488 | struct sock *sk; |
1da177e4 LT |
489 | |
490 | if (!rt) | |
491 | goto out; | |
d8d1f30b | 492 | net = dev_net(rt->dst.dev); |
1da177e4 LT |
493 | |
494 | /* | |
495 | * Find the original header. It is expected to be valid, of course. | |
496 | * Check this, icmp_send is called from the most obscure devices | |
497 | * sometimes. | |
498 | */ | |
eddc9ec5 | 499 | iph = ip_hdr(skb_in); |
1da177e4 | 500 | |
27a884dc ACM |
501 | if ((u8 *)iph < skb_in->head || |
502 | (skb_in->network_header + sizeof(*iph)) > skb_in->tail) | |
1da177e4 LT |
503 | goto out; |
504 | ||
505 | /* | |
506 | * No replies to physical multicast/broadcast | |
507 | */ | |
508 | if (skb_in->pkt_type != PACKET_HOST) | |
509 | goto out; | |
510 | ||
511 | /* | |
512 | * Now check at the protocol level | |
513 | */ | |
514 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
515 | goto out; | |
516 | ||
517 | /* | |
518 | * Only reply to fragment 0. We byte re-order the constant | |
519 | * mask for efficiency. | |
520 | */ | |
521 | if (iph->frag_off & htons(IP_OFFSET)) | |
522 | goto out; | |
523 | ||
524 | /* | |
525 | * If we send an ICMP error to an ICMP error a mess would result.. | |
526 | */ | |
527 | if (icmp_pointers[type].error) { | |
528 | /* | |
529 | * We are an error, check if we are replying to an | |
530 | * ICMP error | |
531 | */ | |
532 | if (iph->protocol == IPPROTO_ICMP) { | |
533 | u8 _inner_type, *itp; | |
534 | ||
535 | itp = skb_header_pointer(skb_in, | |
d56f90a7 | 536 | skb_network_header(skb_in) + |
1da177e4 LT |
537 | (iph->ihl << 2) + |
538 | offsetof(struct icmphdr, | |
539 | type) - | |
540 | skb_in->data, | |
541 | sizeof(_inner_type), | |
542 | &_inner_type); | |
543 | if (itp == NULL) | |
544 | goto out; | |
545 | ||
546 | /* | |
547 | * Assume any unknown ICMP type is an error. This | |
548 | * isn't specified by the RFC, but think about it.. | |
549 | */ | |
550 | if (*itp > NR_ICMP_TYPES || | |
551 | icmp_pointers[*itp].error) | |
552 | goto out; | |
553 | } | |
554 | } | |
555 | ||
fdc0bde9 DL |
556 | sk = icmp_xmit_lock(net); |
557 | if (sk == NULL) | |
1da177e4 LT |
558 | return; |
559 | ||
560 | /* | |
561 | * Construct source address and options. | |
562 | */ | |
563 | ||
564 | saddr = iph->daddr; | |
1c2fb7f9 | 565 | if (!(rt->rt_flags & RTCF_LOCAL)) { |
6e1d9103 PM |
566 | struct net_device *dev = NULL; |
567 | ||
685c7944 | 568 | rcu_read_lock(); |
c7537967 DM |
569 | if (rt_is_input_route(rt) && |
570 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr) | |
5e2b61f7 | 571 | dev = dev_get_by_index_rcu(net, rt->rt_iif); |
6e1d9103 | 572 | |
685c7944 | 573 | if (dev) |
6e1d9103 | 574 | saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
685c7944 | 575 | else |
1c2fb7f9 | 576 | saddr = 0; |
685c7944 | 577 | rcu_read_unlock(); |
1c2fb7f9 | 578 | } |
1da177e4 LT |
579 | |
580 | tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) | | |
581 | IPTOS_PREC_INTERNETCONTROL) : | |
582 | iph->tos; | |
583 | ||
f6d8bd05 | 584 | if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in)) |
fa60cf7f | 585 | goto out_unlock; |
1da177e4 LT |
586 | |
587 | ||
588 | /* | |
589 | * Prepare data for ICMP header. | |
590 | */ | |
591 | ||
592 | icmp_param.data.icmph.type = type; | |
593 | icmp_param.data.icmph.code = code; | |
594 | icmp_param.data.icmph.un.gateway = info; | |
595 | icmp_param.data.icmph.checksum = 0; | |
596 | icmp_param.skb = skb_in; | |
d56f90a7 | 597 | icmp_param.offset = skb_network_offset(skb_in); |
405666db | 598 | inet_sk(sk)->tos = tos; |
1da177e4 | 599 | ipc.addr = iph->saddr; |
f6d8bd05 | 600 | ipc.opt = &icmp_param.replyopts.opt; |
2244d07b | 601 | ipc.tx_flags = 0; |
1da177e4 | 602 | |
77968b78 | 603 | rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, |
f6d460cf DM |
604 | type, code, &icmp_param); |
605 | if (IS_ERR(rt)) | |
606 | goto out_unlock; | |
1da177e4 | 607 | |
a48eff12 | 608 | if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
1da177e4 LT |
609 | goto ende; |
610 | ||
611 | /* RFC says return as much as we can without exceeding 576 bytes. */ | |
612 | ||
d8d1f30b | 613 | room = dst_mtu(&rt->dst); |
1da177e4 LT |
614 | if (room > 576) |
615 | room = 576; | |
f6d8bd05 | 616 | room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen; |
1da177e4 LT |
617 | room -= sizeof(struct icmphdr); |
618 | ||
619 | icmp_param.data_len = skb_in->len - icmp_param.offset; | |
620 | if (icmp_param.data_len > room) | |
621 | icmp_param.data_len = room; | |
622 | icmp_param.head_len = sizeof(struct icmphdr); | |
623 | ||
77968b78 | 624 | icmp_push_reply(&icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
625 | ende: |
626 | ip_rt_put(rt); | |
627 | out_unlock: | |
405666db | 628 | icmp_xmit_unlock(sk); |
1da177e4 LT |
629 | out:; |
630 | } | |
4bc2f18b | 631 | EXPORT_SYMBOL(icmp_send); |
1da177e4 LT |
632 | |
633 | ||
634 | /* | |
635 | * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH. | |
636 | */ | |
637 | ||
638 | static void icmp_unreach(struct sk_buff *skb) | |
639 | { | |
b71d1d42 | 640 | const struct iphdr *iph; |
1da177e4 LT |
641 | struct icmphdr *icmph; |
642 | int hash, protocol; | |
32613090 | 643 | const struct net_protocol *ipprot; |
1da177e4 | 644 | u32 info = 0; |
dde1bc0e DL |
645 | struct net *net; |
646 | ||
adf30907 | 647 | net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
648 | |
649 | /* | |
650 | * Incomplete header ? | |
651 | * Only checks for the IP header, there should be an | |
652 | * additional check for longer headers in upper levels. | |
653 | */ | |
654 | ||
655 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
656 | goto out_err; | |
657 | ||
88c7664f | 658 | icmph = icmp_hdr(skb); |
b71d1d42 | 659 | iph = (const struct iphdr *)skb->data; |
1da177e4 LT |
660 | |
661 | if (iph->ihl < 5) /* Mangled header, drop. */ | |
662 | goto out_err; | |
663 | ||
664 | if (icmph->type == ICMP_DEST_UNREACH) { | |
665 | switch (icmph->code & 15) { | |
666 | case ICMP_NET_UNREACH: | |
667 | case ICMP_HOST_UNREACH: | |
668 | case ICMP_PROT_UNREACH: | |
669 | case ICMP_PORT_UNREACH: | |
670 | break; | |
671 | case ICMP_FRAG_NEEDED: | |
672 | if (ipv4_config.no_pmtu_disc) { | |
afd46503 | 673 | LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"), |
673d57e7 | 674 | &iph->daddr); |
1da177e4 | 675 | } else { |
46517008 | 676 | info = ntohs(icmph->un.frag.mtu); |
1da177e4 LT |
677 | if (!info) |
678 | goto out; | |
679 | } | |
680 | break; | |
681 | case ICMP_SR_FAILED: | |
afd46503 | 682 | LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"), |
673d57e7 | 683 | &iph->daddr); |
1da177e4 LT |
684 | break; |
685 | default: | |
686 | break; | |
687 | } | |
688 | if (icmph->code > NR_ICMP_UNREACH) | |
689 | goto out; | |
690 | } else if (icmph->type == ICMP_PARAMETERPROB) | |
691 | info = ntohl(icmph->un.gateway) >> 24; | |
692 | ||
693 | /* | |
694 | * Throw it at our lower layers | |
695 | * | |
696 | * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed | |
697 | * header. | |
698 | * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the | |
699 | * transport layer. | |
700 | * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to | |
701 | * transport layer. | |
702 | */ | |
703 | ||
704 | /* | |
25985edc | 705 | * Check the other end isn't violating RFC 1122. Some routers send |
1da177e4 LT |
706 | * bogus responses to broadcast frames. If you see this message |
707 | * first check your netmask matches at both ends, if it does then | |
708 | * get the other vendor to fix their kit. | |
709 | */ | |
710 | ||
b34a95ee | 711 | if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses && |
dde1bc0e | 712 | inet_addr_type(net, iph->daddr) == RTN_BROADCAST) { |
e87cc472 JP |
713 | net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n", |
714 | &ip_hdr(skb)->saddr, | |
715 | icmph->type, icmph->code, | |
716 | &iph->daddr, skb->dev->name); | |
1da177e4 LT |
717 | goto out; |
718 | } | |
719 | ||
720 | /* Checkin full IP header plus 8 bytes of protocol to | |
721 | * avoid additional coding at protocol handlers. | |
722 | */ | |
723 | if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) | |
724 | goto out; | |
725 | ||
b71d1d42 | 726 | iph = (const struct iphdr *)skb->data; |
1da177e4 LT |
727 | protocol = iph->protocol; |
728 | ||
729 | /* | |
730 | * Deliver ICMP message to raw sockets. Pretty useless feature? | |
731 | */ | |
7bc54c90 | 732 | raw_icmp_error(skb, protocol, info); |
1da177e4 | 733 | |
1da177e4 | 734 | hash = protocol & (MAX_INET_PROTOS - 1); |
1da177e4 LT |
735 | rcu_read_lock(); |
736 | ipprot = rcu_dereference(inet_protos[hash]); | |
737 | if (ipprot && ipprot->err_handler) | |
738 | ipprot->err_handler(skb, info); | |
739 | rcu_read_unlock(); | |
740 | ||
741 | out: | |
742 | return; | |
743 | out_err: | |
dcfc23ca | 744 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
745 | goto out; |
746 | } | |
747 | ||
748 | ||
749 | /* | |
750 | * Handle ICMP_REDIRECT. | |
751 | */ | |
752 | ||
753 | static void icmp_redirect(struct sk_buff *skb) | |
754 | { | |
b71d1d42 | 755 | const struct iphdr *iph; |
1da177e4 LT |
756 | |
757 | if (skb->len < sizeof(struct iphdr)) | |
758 | goto out_err; | |
759 | ||
760 | /* | |
761 | * Get the copied header of the packet that caused the redirect | |
762 | */ | |
763 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
764 | goto out; | |
765 | ||
b71d1d42 | 766 | iph = (const struct iphdr *)skb->data; |
1da177e4 | 767 | |
88c7664f | 768 | switch (icmp_hdr(skb)->code & 7) { |
1da177e4 LT |
769 | case ICMP_REDIR_NET: |
770 | case ICMP_REDIR_NETTOS: | |
771 | /* | |
772 | * As per RFC recommendations now handle it as a host redirect. | |
773 | */ | |
774 | case ICMP_REDIR_HOST: | |
775 | case ICMP_REDIR_HOSTTOS: | |
eddc9ec5 | 776 | ip_rt_redirect(ip_hdr(skb)->saddr, iph->daddr, |
88c7664f | 777 | icmp_hdr(skb)->un.gateway, |
cef2685e | 778 | iph->saddr, skb->dev); |
1da177e4 | 779 | break; |
e905a9ed | 780 | } |
c319b4d7 VK |
781 | |
782 | /* Ping wants to see redirects. | |
783 | * Let's pretend they are errors of sorts... */ | |
784 | if (iph->protocol == IPPROTO_ICMP && | |
785 | iph->ihl >= 5 && | |
786 | pskb_may_pull(skb, (iph->ihl<<2)+8)) { | |
787 | ping_err(skb, icmp_hdr(skb)->un.gateway); | |
788 | } | |
789 | ||
1da177e4 LT |
790 | out: |
791 | return; | |
792 | out_err: | |
dcfc23ca | 793 | ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS); |
1da177e4 LT |
794 | goto out; |
795 | } | |
796 | ||
797 | /* | |
798 | * Handle ICMP_ECHO ("ping") requests. | |
799 | * | |
800 | * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo | |
801 | * requests. | |
802 | * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be | |
803 | * included in the reply. | |
804 | * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring | |
805 | * echo requests, MUST have default=NOT. | |
806 | * See also WRT handling of options once they are done and working. | |
807 | */ | |
808 | ||
809 | static void icmp_echo(struct sk_buff *skb) | |
810 | { | |
b34a95ee PE |
811 | struct net *net; |
812 | ||
adf30907 | 813 | net = dev_net(skb_dst(skb)->dev); |
b34a95ee | 814 | if (!net->ipv4.sysctl_icmp_echo_ignore_all) { |
1da177e4 LT |
815 | struct icmp_bxm icmp_param; |
816 | ||
88c7664f | 817 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
818 | icmp_param.data.icmph.type = ICMP_ECHOREPLY; |
819 | icmp_param.skb = skb; | |
820 | icmp_param.offset = 0; | |
821 | icmp_param.data_len = skb->len; | |
822 | icmp_param.head_len = sizeof(struct icmphdr); | |
823 | icmp_reply(&icmp_param, skb); | |
824 | } | |
825 | } | |
826 | ||
827 | /* | |
828 | * Handle ICMP Timestamp requests. | |
829 | * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests. | |
830 | * SHOULD be in the kernel for minimum random latency. | |
831 | * MUST be accurate to a few minutes. | |
832 | * MUST be updated at least at 15Hz. | |
833 | */ | |
834 | static void icmp_timestamp(struct sk_buff *skb) | |
835 | { | |
f25c3d61 | 836 | struct timespec tv; |
1da177e4 LT |
837 | struct icmp_bxm icmp_param; |
838 | /* | |
839 | * Too short. | |
840 | */ | |
841 | if (skb->len < 4) | |
842 | goto out_err; | |
843 | ||
844 | /* | |
845 | * Fill in the current time as ms since midnight UT: | |
846 | */ | |
f25c3d61 YH |
847 | getnstimeofday(&tv); |
848 | icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + | |
849 | tv.tv_nsec / NSEC_PER_MSEC); | |
1da177e4 LT |
850 | icmp_param.data.times[2] = icmp_param.data.times[1]; |
851 | if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4)) | |
852 | BUG(); | |
88c7664f | 853 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
854 | icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY; |
855 | icmp_param.data.icmph.code = 0; | |
856 | icmp_param.skb = skb; | |
857 | icmp_param.offset = 0; | |
858 | icmp_param.data_len = 0; | |
859 | icmp_param.head_len = sizeof(struct icmphdr) + 12; | |
860 | icmp_reply(&icmp_param, skb); | |
861 | out: | |
862 | return; | |
863 | out_err: | |
adf30907 | 864 | ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS); |
1da177e4 LT |
865 | goto out; |
866 | } | |
867 | ||
868 | ||
869 | /* | |
870 | * Handle ICMP_ADDRESS_MASK requests. (RFC950) | |
871 | * | |
872 | * RFC1122 (3.2.2.9). A host MUST only send replies to | |
873 | * ADDRESS_MASK requests if it's been configured as an address mask | |
874 | * agent. Receiving a request doesn't constitute implicit permission to | |
875 | * act as one. Of course, implementing this correctly requires (SHOULD) | |
876 | * a way to turn the functionality on and off. Another one for sysctl(), | |
877 | * I guess. -- MS | |
878 | * | |
879 | * RFC1812 (4.3.3.9). A router MUST implement it. | |
880 | * A router SHOULD have switch turning it on/off. | |
881 | * This switch MUST be ON by default. | |
882 | * | |
883 | * Gratuitous replies, zero-source replies are not implemented, | |
884 | * that complies with RFC. DO NOT implement them!!! All the idea | |
885 | * of broadcast addrmask replies as specified in RFC950 is broken. | |
886 | * The problem is that it is not uncommon to have several prefixes | |
887 | * on one physical interface. Moreover, addrmask agent can even be | |
888 | * not aware of existing another prefixes. | |
889 | * If source is zero, addrmask agent cannot choose correct prefix. | |
890 | * Gratuitous mask announcements suffer from the same problem. | |
891 | * RFC1812 explains it, but still allows to use ADDRMASK, | |
892 | * that is pretty silly. --ANK | |
893 | * | |
894 | * All these rules are so bizarre, that I removed kernel addrmask | |
895 | * support at all. It is wrong, it is obsolete, nobody uses it in | |
896 | * any case. --ANK | |
897 | * | |
898 | * Furthermore you can do it with a usermode address agent program | |
899 | * anyway... | |
900 | */ | |
901 | ||
902 | static void icmp_address(struct sk_buff *skb) | |
903 | { | |
904 | #if 0 | |
e87cc472 | 905 | net_dbg_ratelimited("a guy asks for address mask. Who is it?\n"); |
1da177e4 LT |
906 | #endif |
907 | } | |
908 | ||
909 | /* | |
910 | * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain | |
911 | * loudly if an inconsistency is found. | |
cfa087f6 | 912 | * called with rcu_read_lock() |
1da177e4 LT |
913 | */ |
914 | ||
915 | static void icmp_address_reply(struct sk_buff *skb) | |
916 | { | |
511c3f92 | 917 | struct rtable *rt = skb_rtable(skb); |
1da177e4 LT |
918 | struct net_device *dev = skb->dev; |
919 | struct in_device *in_dev; | |
920 | struct in_ifaddr *ifa; | |
921 | ||
922 | if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC)) | |
cfa087f6 | 923 | return; |
1da177e4 | 924 | |
cfa087f6 | 925 | in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 926 | if (!in_dev) |
cfa087f6 ED |
927 | return; |
928 | ||
1da177e4 LT |
929 | if (in_dev->ifa_list && |
930 | IN_DEV_LOG_MARTIANS(in_dev) && | |
931 | IN_DEV_FORWARD(in_dev)) { | |
a144ea4b | 932 | __be32 _mask, *mp; |
1da177e4 LT |
933 | |
934 | mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask); | |
09a62660 | 935 | BUG_ON(mp == NULL); |
1da177e4 LT |
936 | for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { |
937 | if (*mp == ifa->ifa_mask && | |
9f6abb5f | 938 | inet_ifa_match(ip_hdr(skb)->saddr, ifa)) |
1da177e4 LT |
939 | break; |
940 | } | |
e87cc472 JP |
941 | if (!ifa) |
942 | net_info_ratelimited("Wrong address mask %pI4 from %s/%pI4\n", | |
943 | mp, | |
944 | dev->name, &ip_hdr(skb)->saddr); | |
1da177e4 | 945 | } |
1da177e4 LT |
946 | } |
947 | ||
948 | static void icmp_discard(struct sk_buff *skb) | |
949 | { | |
950 | } | |
951 | ||
952 | /* | |
953 | * Deal with incoming ICMP packets. | |
954 | */ | |
955 | int icmp_rcv(struct sk_buff *skb) | |
956 | { | |
957 | struct icmphdr *icmph; | |
511c3f92 | 958 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 959 | struct net *net = dev_net(rt->dst.dev); |
1da177e4 | 960 | |
aebcf82c | 961 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { |
def8b4fa | 962 | struct sec_path *sp = skb_sec_path(skb); |
8b7817f3 HX |
963 | int nh; |
964 | ||
def8b4fa | 965 | if (!(sp && sp->xvec[sp->len - 1]->props.flags & |
aebcf82c HX |
966 | XFRM_STATE_ICMP)) |
967 | goto drop; | |
968 | ||
8b7817f3 HX |
969 | if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr))) |
970 | goto drop; | |
971 | ||
972 | nh = skb_network_offset(skb); | |
973 | skb_set_network_header(skb, sizeof(*icmph)); | |
974 | ||
975 | if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) | |
976 | goto drop; | |
977 | ||
978 | skb_set_network_header(skb, nh); | |
979 | } | |
980 | ||
dcfc23ca | 981 | ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS); |
1da177e4 LT |
982 | |
983 | switch (skb->ip_summed) { | |
84fa7933 | 984 | case CHECKSUM_COMPLETE: |
d3bc23e7 | 985 | if (!csum_fold(skb->csum)) |
1da177e4 | 986 | break; |
fb286bb2 | 987 | /* fall through */ |
1da177e4 | 988 | case CHECKSUM_NONE: |
fb286bb2 HX |
989 | skb->csum = 0; |
990 | if (__skb_checksum_complete(skb)) | |
1da177e4 | 991 | goto error; |
1da177e4 LT |
992 | } |
993 | ||
8cf22943 HX |
994 | if (!pskb_pull(skb, sizeof(*icmph))) |
995 | goto error; | |
1da177e4 | 996 | |
88c7664f | 997 | icmph = icmp_hdr(skb); |
1da177e4 | 998 | |
f66ac03d | 999 | ICMPMSGIN_INC_STATS_BH(net, icmph->type); |
1da177e4 LT |
1000 | /* |
1001 | * 18 is the highest 'known' ICMP type. Anything else is a mystery | |
1002 | * | |
1003 | * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently | |
1004 | * discarded. | |
1005 | */ | |
1006 | if (icmph->type > NR_ICMP_TYPES) | |
1007 | goto error; | |
1008 | ||
1009 | ||
1010 | /* | |
1011 | * Parse the ICMP message | |
1012 | */ | |
1013 | ||
e905a9ed | 1014 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
1da177e4 LT |
1015 | /* |
1016 | * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be | |
1017 | * silently ignored (we let user decide with a sysctl). | |
1018 | * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently | |
1019 | * discarded if to broadcast/multicast. | |
1020 | */ | |
4c866aa7 AK |
1021 | if ((icmph->type == ICMP_ECHO || |
1022 | icmph->type == ICMP_TIMESTAMP) && | |
b34a95ee | 1023 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts) { |
1da177e4 LT |
1024 | goto error; |
1025 | } | |
1026 | if (icmph->type != ICMP_ECHO && | |
1027 | icmph->type != ICMP_TIMESTAMP && | |
1028 | icmph->type != ICMP_ADDRESS && | |
1029 | icmph->type != ICMP_ADDRESSREPLY) { | |
1030 | goto error; | |
e905a9ed | 1031 | } |
1da177e4 LT |
1032 | } |
1033 | ||
1da177e4 LT |
1034 | icmp_pointers[icmph->type].handler(skb); |
1035 | ||
1036 | drop: | |
1037 | kfree_skb(skb); | |
1038 | return 0; | |
1039 | error: | |
dcfc23ca | 1040 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
1041 | goto drop; |
1042 | } | |
1043 | ||
1044 | /* | |
1045 | * This table is the definition of how we handle ICMP. | |
1046 | */ | |
9b5b5cff | 1047 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = { |
1da177e4 | 1048 | [ICMP_ECHOREPLY] = { |
c319b4d7 | 1049 | .handler = ping_rcv, |
1da177e4 LT |
1050 | }, |
1051 | [1] = { | |
1da177e4 LT |
1052 | .handler = icmp_discard, |
1053 | .error = 1, | |
1054 | }, | |
1055 | [2] = { | |
1da177e4 LT |
1056 | .handler = icmp_discard, |
1057 | .error = 1, | |
1058 | }, | |
1059 | [ICMP_DEST_UNREACH] = { | |
1da177e4 LT |
1060 | .handler = icmp_unreach, |
1061 | .error = 1, | |
1062 | }, | |
1063 | [ICMP_SOURCE_QUENCH] = { | |
1da177e4 LT |
1064 | .handler = icmp_unreach, |
1065 | .error = 1, | |
1066 | }, | |
1067 | [ICMP_REDIRECT] = { | |
1da177e4 LT |
1068 | .handler = icmp_redirect, |
1069 | .error = 1, | |
1070 | }, | |
1071 | [6] = { | |
1da177e4 LT |
1072 | .handler = icmp_discard, |
1073 | .error = 1, | |
1074 | }, | |
1075 | [7] = { | |
1da177e4 LT |
1076 | .handler = icmp_discard, |
1077 | .error = 1, | |
1078 | }, | |
1079 | [ICMP_ECHO] = { | |
1da177e4 LT |
1080 | .handler = icmp_echo, |
1081 | }, | |
1082 | [9] = { | |
1da177e4 LT |
1083 | .handler = icmp_discard, |
1084 | .error = 1, | |
1085 | }, | |
1086 | [10] = { | |
1da177e4 LT |
1087 | .handler = icmp_discard, |
1088 | .error = 1, | |
1089 | }, | |
1090 | [ICMP_TIME_EXCEEDED] = { | |
1da177e4 LT |
1091 | .handler = icmp_unreach, |
1092 | .error = 1, | |
1093 | }, | |
1094 | [ICMP_PARAMETERPROB] = { | |
1da177e4 LT |
1095 | .handler = icmp_unreach, |
1096 | .error = 1, | |
1097 | }, | |
1098 | [ICMP_TIMESTAMP] = { | |
1da177e4 LT |
1099 | .handler = icmp_timestamp, |
1100 | }, | |
1101 | [ICMP_TIMESTAMPREPLY] = { | |
1da177e4 LT |
1102 | .handler = icmp_discard, |
1103 | }, | |
1104 | [ICMP_INFO_REQUEST] = { | |
1da177e4 LT |
1105 | .handler = icmp_discard, |
1106 | }, | |
e905a9ed | 1107 | [ICMP_INFO_REPLY] = { |
1da177e4 LT |
1108 | .handler = icmp_discard, |
1109 | }, | |
1110 | [ICMP_ADDRESS] = { | |
1da177e4 LT |
1111 | .handler = icmp_address, |
1112 | }, | |
1113 | [ICMP_ADDRESSREPLY] = { | |
1da177e4 LT |
1114 | .handler = icmp_address_reply, |
1115 | }, | |
1116 | }; | |
1117 | ||
4a6ad7a1 | 1118 | static void __net_exit icmp_sk_exit(struct net *net) |
1da177e4 | 1119 | { |
1da177e4 LT |
1120 | int i; |
1121 | ||
5c8cafd6 | 1122 | for_each_possible_cpu(i) |
c1e9894d | 1123 | inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]); |
4a6ad7a1 DL |
1124 | kfree(net->ipv4.icmp_sk); |
1125 | net->ipv4.icmp_sk = NULL; | |
a5710d65 | 1126 | } |
1da177e4 | 1127 | |
263173af | 1128 | static int __net_init icmp_sk_init(struct net *net) |
a5710d65 | 1129 | { |
a5710d65 DL |
1130 | int i, err; |
1131 | ||
4a6ad7a1 DL |
1132 | net->ipv4.icmp_sk = |
1133 | kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL); | |
1134 | if (net->ipv4.icmp_sk == NULL) | |
79c91159 DL |
1135 | return -ENOMEM; |
1136 | ||
a5710d65 | 1137 | for_each_possible_cpu(i) { |
1e3cf683 | 1138 | struct sock *sk; |
1da177e4 | 1139 | |
c1e9894d DL |
1140 | err = inet_ctl_sock_create(&sk, PF_INET, |
1141 | SOCK_RAW, IPPROTO_ICMP, net); | |
1da177e4 | 1142 | if (err < 0) |
a5710d65 | 1143 | goto fail; |
1da177e4 | 1144 | |
c1e9894d | 1145 | net->ipv4.icmp_sk[i] = sk; |
1da177e4 LT |
1146 | |
1147 | /* Enough space for 2 64K ICMP packets, including | |
87fb4b7b | 1148 | * sk_buff/skb_shared_info struct overhead. |
1da177e4 | 1149 | */ |
87fb4b7b | 1150 | sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); |
1da177e4 | 1151 | |
b3a5b6cc ED |
1152 | /* |
1153 | * Speedup sock_wfree() | |
1154 | */ | |
1155 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
c1e9894d | 1156 | inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT; |
1da177e4 | 1157 | } |
a24022e1 PE |
1158 | |
1159 | /* Control parameters for ECHO replies. */ | |
1160 | net->ipv4.sysctl_icmp_echo_ignore_all = 0; | |
1161 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1; | |
1162 | ||
1163 | /* Control parameter - ignore bogus broadcast responses? */ | |
1164 | net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1; | |
1165 | ||
1166 | /* | |
1167 | * Configurable global rate limit. | |
1168 | * | |
1169 | * ratelimit defines tokens/packet consumed for dst->rate_token | |
1170 | * bucket ratemask defines which icmp types are ratelimited by | |
1171 | * setting it's bit position. | |
1172 | * | |
1173 | * default: | |
1174 | * dest unreachable (3), source quench (4), | |
1175 | * time exceeded (11), parameter problem (12) | |
1176 | */ | |
1177 | ||
1178 | net->ipv4.sysctl_icmp_ratelimit = 1 * HZ; | |
1179 | net->ipv4.sysctl_icmp_ratemask = 0x1818; | |
1180 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0; | |
1181 | ||
a5710d65 DL |
1182 | return 0; |
1183 | ||
1184 | fail: | |
1d1c8d13 | 1185 | for_each_possible_cpu(i) |
c1e9894d | 1186 | inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]); |
1d1c8d13 | 1187 | kfree(net->ipv4.icmp_sk); |
a5710d65 | 1188 | return err; |
1da177e4 LT |
1189 | } |
1190 | ||
4a6ad7a1 DL |
1191 | static struct pernet_operations __net_initdata icmp_sk_ops = { |
1192 | .init = icmp_sk_init, | |
1193 | .exit = icmp_sk_exit, | |
1194 | }; | |
1195 | ||
1196 | int __init icmp_init(void) | |
1197 | { | |
959d2726 | 1198 | return register_pernet_subsys(&icmp_sk_ops); |
4a6ad7a1 | 1199 | } |