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); |
1d861aa4 | 260 | inet_putpeer(peer); |
92d86829 | 261 | } |
1da177e4 LT |
262 | out: |
263 | return rc; | |
264 | } | |
265 | ||
266 | /* | |
267 | * Maintain the counters used in the SNMP statistics for outgoing ICMP | |
268 | */ | |
0388b004 | 269 | void icmp_out_count(struct net *net, unsigned char type) |
1da177e4 | 270 | { |
903fc196 | 271 | ICMPMSGOUT_INC_STATS(net, type); |
75c939bb | 272 | ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS); |
1da177e4 LT |
273 | } |
274 | ||
275 | /* | |
276 | * Checksum each fragment, and on the first include the headers and final | |
277 | * checksum. | |
278 | */ | |
279 | static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd, | |
280 | struct sk_buff *skb) | |
281 | { | |
282 | struct icmp_bxm *icmp_param = (struct icmp_bxm *)from; | |
5f92a738 | 283 | __wsum csum; |
1da177e4 LT |
284 | |
285 | csum = skb_copy_and_csum_bits(icmp_param->skb, | |
286 | icmp_param->offset + offset, | |
287 | to, len, 0); | |
288 | ||
289 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
290 | if (icmp_pointers[icmp_param->data.icmph.type].error) | |
291 | nf_ct_attach(skb, icmp_param->skb); | |
292 | return 0; | |
293 | } | |
294 | ||
295 | static void icmp_push_reply(struct icmp_bxm *icmp_param, | |
77968b78 | 296 | struct flowi4 *fl4, |
2e77d89b | 297 | struct ipcm_cookie *ipc, struct rtable **rt) |
1da177e4 | 298 | { |
1e3cf683 | 299 | struct sock *sk; |
1da177e4 LT |
300 | struct sk_buff *skb; |
301 | ||
d8d1f30b | 302 | sk = icmp_sk(dev_net((*rt)->dst.dev)); |
f5fca608 | 303 | if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param, |
e905a9ed YH |
304 | icmp_param->data_len+icmp_param->head_len, |
305 | icmp_param->head_len, | |
1f8438a8 ED |
306 | ipc, rt, MSG_DONTWAIT) < 0) { |
307 | ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS); | |
1e3cf683 | 308 | ip_flush_pending_frames(sk); |
1f8438a8 | 309 | } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) { |
88c7664f | 310 | struct icmphdr *icmph = icmp_hdr(skb); |
d3bc23e7 | 311 | __wsum csum = 0; |
1da177e4 LT |
312 | struct sk_buff *skb1; |
313 | ||
1e3cf683 | 314 | skb_queue_walk(&sk->sk_write_queue, skb1) { |
1da177e4 LT |
315 | csum = csum_add(csum, skb1->csum); |
316 | } | |
317 | csum = csum_partial_copy_nocheck((void *)&icmp_param->data, | |
318 | (char *)icmph, | |
319 | icmp_param->head_len, csum); | |
320 | icmph->checksum = csum_fold(csum); | |
321 | skb->ip_summed = CHECKSUM_NONE; | |
77968b78 | 322 | ip_push_pending_frames(sk, fl4); |
1da177e4 LT |
323 | } |
324 | } | |
325 | ||
326 | /* | |
327 | * Driving logic for building and sending ICMP messages. | |
328 | */ | |
329 | ||
330 | static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb) | |
331 | { | |
1da177e4 | 332 | struct ipcm_cookie ipc; |
511c3f92 | 333 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 334 | struct net *net = dev_net(rt->dst.dev); |
77968b78 | 335 | struct flowi4 fl4; |
fdc0bde9 DL |
336 | struct sock *sk; |
337 | struct inet_sock *inet; | |
35ebf65e | 338 | __be32 daddr, saddr; |
1da177e4 | 339 | |
f6d8bd05 | 340 | if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb)) |
f00c401b | 341 | return; |
1da177e4 | 342 | |
fdc0bde9 DL |
343 | sk = icmp_xmit_lock(net); |
344 | if (sk == NULL) | |
1da177e4 | 345 | return; |
fdc0bde9 | 346 | inet = inet_sk(sk); |
1da177e4 LT |
347 | |
348 | icmp_param->data.icmph.checksum = 0; | |
1da177e4 | 349 | |
eddc9ec5 | 350 | inet->tos = ip_hdr(skb)->tos; |
9f6abb5f | 351 | daddr = ipc.addr = ip_hdr(skb)->saddr; |
35ebf65e | 352 | saddr = fib_compute_spec_dst(skb); |
1da177e4 | 353 | ipc.opt = NULL; |
2244d07b | 354 | ipc.tx_flags = 0; |
f6d8bd05 ED |
355 | if (icmp_param->replyopts.opt.opt.optlen) { |
356 | ipc.opt = &icmp_param->replyopts.opt; | |
357 | if (ipc.opt->opt.srr) | |
358 | daddr = icmp_param->replyopts.opt.opt.faddr; | |
1da177e4 | 359 | } |
77968b78 DM |
360 | memset(&fl4, 0, sizeof(fl4)); |
361 | fl4.daddr = daddr; | |
35ebf65e | 362 | fl4.saddr = saddr; |
77968b78 DM |
363 | fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos); |
364 | fl4.flowi4_proto = IPPROTO_ICMP; | |
365 | security_skb_classify_flow(skb, flowi4_to_flowi(&fl4)); | |
366 | rt = ip_route_output_key(net, &fl4); | |
367 | if (IS_ERR(rt)) | |
368 | goto out_unlock; | |
a48eff12 | 369 | if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type, |
1da177e4 | 370 | icmp_param->data.icmph.code)) |
77968b78 | 371 | icmp_push_reply(icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
372 | ip_rt_put(rt); |
373 | out_unlock: | |
405666db | 374 | icmp_xmit_unlock(sk); |
1da177e4 LT |
375 | } |
376 | ||
77968b78 DM |
377 | static struct rtable *icmp_route_lookup(struct net *net, |
378 | struct flowi4 *fl4, | |
379 | struct sk_buff *skb_in, | |
b71d1d42 | 380 | const struct iphdr *iph, |
f6d460cf DM |
381 | __be32 saddr, u8 tos, |
382 | int type, int code, | |
383 | struct icmp_bxm *param) | |
384 | { | |
f6d460cf | 385 | struct rtable *rt, *rt2; |
415b3334 | 386 | struct flowi4 fl4_dec; |
f6d460cf DM |
387 | int err; |
388 | ||
77968b78 DM |
389 | memset(fl4, 0, sizeof(*fl4)); |
390 | fl4->daddr = (param->replyopts.opt.opt.srr ? | |
391 | param->replyopts.opt.opt.faddr : iph->saddr); | |
392 | fl4->saddr = saddr; | |
393 | fl4->flowi4_tos = RT_TOS(tos); | |
394 | fl4->flowi4_proto = IPPROTO_ICMP; | |
395 | fl4->fl4_icmp_type = type; | |
396 | fl4->fl4_icmp_code = code; | |
397 | security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4)); | |
398 | rt = __ip_route_output_key(net, fl4); | |
b23dd4fe DM |
399 | if (IS_ERR(rt)) |
400 | return rt; | |
f6d460cf DM |
401 | |
402 | /* No need to clone since we're just using its address. */ | |
403 | rt2 = rt; | |
404 | ||
9d6ec938 | 405 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
77968b78 | 406 | flowi4_to_flowi(fl4), NULL, 0); |
452edd59 | 407 | if (!IS_ERR(rt)) { |
f6d460cf DM |
408 | if (rt != rt2) |
409 | return rt; | |
452edd59 | 410 | } else if (PTR_ERR(rt) == -EPERM) { |
f6d460cf | 411 | rt = NULL; |
452edd59 DM |
412 | } else |
413 | return rt; | |
f6d460cf | 414 | |
415b3334 | 415 | err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET); |
f6d460cf DM |
416 | if (err) |
417 | goto relookup_failed; | |
418 | ||
415b3334 DM |
419 | if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) { |
420 | rt2 = __ip_route_output_key(net, &fl4_dec); | |
b23dd4fe DM |
421 | if (IS_ERR(rt2)) |
422 | err = PTR_ERR(rt2); | |
f6d460cf | 423 | } else { |
9d6ec938 | 424 | struct flowi4 fl4_2 = {}; |
f6d460cf DM |
425 | unsigned long orefdst; |
426 | ||
415b3334 | 427 | fl4_2.daddr = fl4_dec.saddr; |
9d6ec938 | 428 | rt2 = ip_route_output_key(net, &fl4_2); |
b23dd4fe DM |
429 | if (IS_ERR(rt2)) { |
430 | err = PTR_ERR(rt2); | |
f6d460cf | 431 | goto relookup_failed; |
b23dd4fe | 432 | } |
f6d460cf DM |
433 | /* Ugh! */ |
434 | orefdst = skb_in->_skb_refdst; /* save old refdst */ | |
415b3334 | 435 | err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr, |
f6d460cf DM |
436 | RT_TOS(tos), rt2->dst.dev); |
437 | ||
438 | dst_release(&rt2->dst); | |
439 | rt2 = skb_rtable(skb_in); | |
440 | skb_in->_skb_refdst = orefdst; /* restore old refdst */ | |
441 | } | |
442 | ||
443 | if (err) | |
444 | goto relookup_failed; | |
445 | ||
9d6ec938 | 446 | rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst, |
415b3334 | 447 | flowi4_to_flowi(&fl4_dec), NULL, |
9d6ec938 | 448 | XFRM_LOOKUP_ICMP); |
452edd59 | 449 | if (!IS_ERR(rt2)) { |
f6d460cf | 450 | dst_release(&rt->dst); |
415b3334 | 451 | memcpy(fl4, &fl4_dec, sizeof(*fl4)); |
f6d460cf | 452 | rt = rt2; |
452edd59 DM |
453 | } else if (PTR_ERR(rt2) == -EPERM) { |
454 | if (rt) | |
455 | dst_release(&rt->dst); | |
456 | return rt2; | |
457 | } else { | |
458 | err = PTR_ERR(rt2); | |
459 | goto relookup_failed; | |
f6d460cf | 460 | } |
f6d460cf DM |
461 | return rt; |
462 | ||
463 | relookup_failed: | |
464 | if (rt) | |
465 | return rt; | |
466 | return ERR_PTR(err); | |
467 | } | |
1da177e4 LT |
468 | |
469 | /* | |
470 | * Send an ICMP message in response to a situation | |
471 | * | |
472 | * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header. | |
473 | * MAY send more (we do). | |
474 | * MUST NOT change this header information. | |
475 | * MUST NOT reply to a multicast/broadcast IP address. | |
476 | * MUST NOT reply to a multicast/broadcast MAC address. | |
477 | * MUST reply to only the first fragment. | |
478 | */ | |
479 | ||
e4883014 | 480 | void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info) |
1da177e4 LT |
481 | { |
482 | struct iphdr *iph; | |
483 | int room; | |
484 | struct icmp_bxm icmp_param; | |
511c3f92 | 485 | struct rtable *rt = skb_rtable(skb_in); |
1da177e4 | 486 | struct ipcm_cookie ipc; |
77968b78 | 487 | struct flowi4 fl4; |
a61ced5d | 488 | __be32 saddr; |
1da177e4 | 489 | u8 tos; |
dde1bc0e | 490 | struct net *net; |
4a6ad7a1 | 491 | struct sock *sk; |
1da177e4 LT |
492 | |
493 | if (!rt) | |
494 | goto out; | |
d8d1f30b | 495 | net = dev_net(rt->dst.dev); |
1da177e4 LT |
496 | |
497 | /* | |
498 | * Find the original header. It is expected to be valid, of course. | |
499 | * Check this, icmp_send is called from the most obscure devices | |
500 | * sometimes. | |
501 | */ | |
eddc9ec5 | 502 | iph = ip_hdr(skb_in); |
1da177e4 | 503 | |
27a884dc ACM |
504 | if ((u8 *)iph < skb_in->head || |
505 | (skb_in->network_header + sizeof(*iph)) > skb_in->tail) | |
1da177e4 LT |
506 | goto out; |
507 | ||
508 | /* | |
509 | * No replies to physical multicast/broadcast | |
510 | */ | |
511 | if (skb_in->pkt_type != PACKET_HOST) | |
512 | goto out; | |
513 | ||
514 | /* | |
515 | * Now check at the protocol level | |
516 | */ | |
517 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
518 | goto out; | |
519 | ||
520 | /* | |
521 | * Only reply to fragment 0. We byte re-order the constant | |
522 | * mask for efficiency. | |
523 | */ | |
524 | if (iph->frag_off & htons(IP_OFFSET)) | |
525 | goto out; | |
526 | ||
527 | /* | |
528 | * If we send an ICMP error to an ICMP error a mess would result.. | |
529 | */ | |
530 | if (icmp_pointers[type].error) { | |
531 | /* | |
532 | * We are an error, check if we are replying to an | |
533 | * ICMP error | |
534 | */ | |
535 | if (iph->protocol == IPPROTO_ICMP) { | |
536 | u8 _inner_type, *itp; | |
537 | ||
538 | itp = skb_header_pointer(skb_in, | |
d56f90a7 | 539 | skb_network_header(skb_in) + |
1da177e4 LT |
540 | (iph->ihl << 2) + |
541 | offsetof(struct icmphdr, | |
542 | type) - | |
543 | skb_in->data, | |
544 | sizeof(_inner_type), | |
545 | &_inner_type); | |
546 | if (itp == NULL) | |
547 | goto out; | |
548 | ||
549 | /* | |
550 | * Assume any unknown ICMP type is an error. This | |
551 | * isn't specified by the RFC, but think about it.. | |
552 | */ | |
553 | if (*itp > NR_ICMP_TYPES || | |
554 | icmp_pointers[*itp].error) | |
555 | goto out; | |
556 | } | |
557 | } | |
558 | ||
fdc0bde9 DL |
559 | sk = icmp_xmit_lock(net); |
560 | if (sk == NULL) | |
1da177e4 LT |
561 | return; |
562 | ||
563 | /* | |
564 | * Construct source address and options. | |
565 | */ | |
566 | ||
567 | saddr = iph->daddr; | |
1c2fb7f9 | 568 | if (!(rt->rt_flags & RTCF_LOCAL)) { |
6e1d9103 PM |
569 | struct net_device *dev = NULL; |
570 | ||
685c7944 | 571 | rcu_read_lock(); |
c7537967 DM |
572 | if (rt_is_input_route(rt) && |
573 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr) | |
5e2b61f7 | 574 | dev = dev_get_by_index_rcu(net, rt->rt_iif); |
6e1d9103 | 575 | |
685c7944 | 576 | if (dev) |
6e1d9103 | 577 | saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
685c7944 | 578 | else |
1c2fb7f9 | 579 | saddr = 0; |
685c7944 | 580 | rcu_read_unlock(); |
1c2fb7f9 | 581 | } |
1da177e4 LT |
582 | |
583 | tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) | | |
584 | IPTOS_PREC_INTERNETCONTROL) : | |
585 | iph->tos; | |
586 | ||
f6d8bd05 | 587 | if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in)) |
fa60cf7f | 588 | goto out_unlock; |
1da177e4 LT |
589 | |
590 | ||
591 | /* | |
592 | * Prepare data for ICMP header. | |
593 | */ | |
594 | ||
595 | icmp_param.data.icmph.type = type; | |
596 | icmp_param.data.icmph.code = code; | |
597 | icmp_param.data.icmph.un.gateway = info; | |
598 | icmp_param.data.icmph.checksum = 0; | |
599 | icmp_param.skb = skb_in; | |
d56f90a7 | 600 | icmp_param.offset = skb_network_offset(skb_in); |
405666db | 601 | inet_sk(sk)->tos = tos; |
1da177e4 | 602 | ipc.addr = iph->saddr; |
f6d8bd05 | 603 | ipc.opt = &icmp_param.replyopts.opt; |
2244d07b | 604 | ipc.tx_flags = 0; |
1da177e4 | 605 | |
77968b78 | 606 | rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, |
f6d460cf DM |
607 | type, code, &icmp_param); |
608 | if (IS_ERR(rt)) | |
609 | goto out_unlock; | |
1da177e4 | 610 | |
a48eff12 | 611 | if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
1da177e4 LT |
612 | goto ende; |
613 | ||
614 | /* RFC says return as much as we can without exceeding 576 bytes. */ | |
615 | ||
d8d1f30b | 616 | room = dst_mtu(&rt->dst); |
1da177e4 LT |
617 | if (room > 576) |
618 | room = 576; | |
f6d8bd05 | 619 | room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen; |
1da177e4 LT |
620 | room -= sizeof(struct icmphdr); |
621 | ||
622 | icmp_param.data_len = skb_in->len - icmp_param.offset; | |
623 | if (icmp_param.data_len > room) | |
624 | icmp_param.data_len = room; | |
625 | icmp_param.head_len = sizeof(struct icmphdr); | |
626 | ||
77968b78 | 627 | icmp_push_reply(&icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
628 | ende: |
629 | ip_rt_put(rt); | |
630 | out_unlock: | |
405666db | 631 | icmp_xmit_unlock(sk); |
1da177e4 LT |
632 | out:; |
633 | } | |
4bc2f18b | 634 | EXPORT_SYMBOL(icmp_send); |
1da177e4 LT |
635 | |
636 | ||
1de9243b DM |
637 | static void icmp_socket_deliver(struct sk_buff *skb, u32 info) |
638 | { | |
639 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
640 | const struct net_protocol *ipprot; | |
641 | int protocol = iph->protocol; | |
642 | ||
f0a70e90 DM |
643 | /* Checkin full IP header plus 8 bytes of protocol to |
644 | * avoid additional coding at protocol handlers. | |
645 | */ | |
646 | if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) | |
647 | return; | |
648 | ||
1de9243b DM |
649 | raw_icmp_error(skb, protocol, info); |
650 | ||
651 | rcu_read_lock(); | |
652 | ipprot = rcu_dereference(inet_protos[protocol]); | |
653 | if (ipprot && ipprot->err_handler) | |
654 | ipprot->err_handler(skb, info); | |
655 | rcu_read_unlock(); | |
656 | } | |
657 | ||
1da177e4 LT |
658 | /* |
659 | * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH. | |
660 | */ | |
661 | ||
662 | static void icmp_unreach(struct sk_buff *skb) | |
663 | { | |
b71d1d42 | 664 | const struct iphdr *iph; |
1da177e4 | 665 | struct icmphdr *icmph; |
dde1bc0e | 666 | struct net *net; |
f9242b6b | 667 | u32 info = 0; |
dde1bc0e | 668 | |
adf30907 | 669 | net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
670 | |
671 | /* | |
672 | * Incomplete header ? | |
673 | * Only checks for the IP header, there should be an | |
674 | * additional check for longer headers in upper levels. | |
675 | */ | |
676 | ||
677 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
678 | goto out_err; | |
679 | ||
88c7664f | 680 | icmph = icmp_hdr(skb); |
b71d1d42 | 681 | iph = (const struct iphdr *)skb->data; |
1da177e4 LT |
682 | |
683 | if (iph->ihl < 5) /* Mangled header, drop. */ | |
684 | goto out_err; | |
685 | ||
686 | if (icmph->type == ICMP_DEST_UNREACH) { | |
687 | switch (icmph->code & 15) { | |
688 | case ICMP_NET_UNREACH: | |
689 | case ICMP_HOST_UNREACH: | |
690 | case ICMP_PROT_UNREACH: | |
691 | case ICMP_PORT_UNREACH: | |
692 | break; | |
693 | case ICMP_FRAG_NEEDED: | |
694 | if (ipv4_config.no_pmtu_disc) { | |
afd46503 | 695 | LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"), |
673d57e7 | 696 | &iph->daddr); |
1da177e4 | 697 | } else { |
46517008 | 698 | info = ntohs(icmph->un.frag.mtu); |
1da177e4 LT |
699 | if (!info) |
700 | goto out; | |
701 | } | |
702 | break; | |
703 | case ICMP_SR_FAILED: | |
afd46503 | 704 | LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"), |
673d57e7 | 705 | &iph->daddr); |
1da177e4 LT |
706 | break; |
707 | default: | |
708 | break; | |
709 | } | |
710 | if (icmph->code > NR_ICMP_UNREACH) | |
711 | goto out; | |
712 | } else if (icmph->type == ICMP_PARAMETERPROB) | |
713 | info = ntohl(icmph->un.gateway) >> 24; | |
714 | ||
715 | /* | |
716 | * Throw it at our lower layers | |
717 | * | |
718 | * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed | |
719 | * header. | |
720 | * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the | |
721 | * transport layer. | |
722 | * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to | |
723 | * transport layer. | |
724 | */ | |
725 | ||
726 | /* | |
25985edc | 727 | * Check the other end isn't violating RFC 1122. Some routers send |
1da177e4 LT |
728 | * bogus responses to broadcast frames. If you see this message |
729 | * first check your netmask matches at both ends, if it does then | |
730 | * get the other vendor to fix their kit. | |
731 | */ | |
732 | ||
b34a95ee | 733 | if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses && |
dde1bc0e | 734 | inet_addr_type(net, iph->daddr) == RTN_BROADCAST) { |
e87cc472 JP |
735 | net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n", |
736 | &ip_hdr(skb)->saddr, | |
737 | icmph->type, icmph->code, | |
738 | &iph->daddr, skb->dev->name); | |
1da177e4 LT |
739 | goto out; |
740 | } | |
741 | ||
1de9243b | 742 | icmp_socket_deliver(skb, info); |
1da177e4 LT |
743 | |
744 | out: | |
745 | return; | |
746 | out_err: | |
dcfc23ca | 747 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
748 | goto out; |
749 | } | |
750 | ||
751 | ||
752 | /* | |
753 | * Handle ICMP_REDIRECT. | |
754 | */ | |
755 | ||
756 | static void icmp_redirect(struct sk_buff *skb) | |
757 | { | |
94206125 DM |
758 | if (skb->len < sizeof(struct iphdr)) { |
759 | ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS); | |
760 | return; | |
761 | } | |
1da177e4 | 762 | |
1da177e4 | 763 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
94206125 | 764 | return; |
c319b4d7 | 765 | |
d3351b75 | 766 | icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway); |
1da177e4 LT |
767 | } |
768 | ||
769 | /* | |
770 | * Handle ICMP_ECHO ("ping") requests. | |
771 | * | |
772 | * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo | |
773 | * requests. | |
774 | * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be | |
775 | * included in the reply. | |
776 | * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring | |
777 | * echo requests, MUST have default=NOT. | |
778 | * See also WRT handling of options once they are done and working. | |
779 | */ | |
780 | ||
781 | static void icmp_echo(struct sk_buff *skb) | |
782 | { | |
b34a95ee PE |
783 | struct net *net; |
784 | ||
adf30907 | 785 | net = dev_net(skb_dst(skb)->dev); |
b34a95ee | 786 | if (!net->ipv4.sysctl_icmp_echo_ignore_all) { |
1da177e4 LT |
787 | struct icmp_bxm icmp_param; |
788 | ||
88c7664f | 789 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
790 | icmp_param.data.icmph.type = ICMP_ECHOREPLY; |
791 | icmp_param.skb = skb; | |
792 | icmp_param.offset = 0; | |
793 | icmp_param.data_len = skb->len; | |
794 | icmp_param.head_len = sizeof(struct icmphdr); | |
795 | icmp_reply(&icmp_param, skb); | |
796 | } | |
797 | } | |
798 | ||
799 | /* | |
800 | * Handle ICMP Timestamp requests. | |
801 | * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests. | |
802 | * SHOULD be in the kernel for minimum random latency. | |
803 | * MUST be accurate to a few minutes. | |
804 | * MUST be updated at least at 15Hz. | |
805 | */ | |
806 | static void icmp_timestamp(struct sk_buff *skb) | |
807 | { | |
f25c3d61 | 808 | struct timespec tv; |
1da177e4 LT |
809 | struct icmp_bxm icmp_param; |
810 | /* | |
811 | * Too short. | |
812 | */ | |
813 | if (skb->len < 4) | |
814 | goto out_err; | |
815 | ||
816 | /* | |
817 | * Fill in the current time as ms since midnight UT: | |
818 | */ | |
f25c3d61 YH |
819 | getnstimeofday(&tv); |
820 | icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + | |
821 | tv.tv_nsec / NSEC_PER_MSEC); | |
1da177e4 LT |
822 | icmp_param.data.times[2] = icmp_param.data.times[1]; |
823 | if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4)) | |
824 | BUG(); | |
88c7664f | 825 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
826 | icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY; |
827 | icmp_param.data.icmph.code = 0; | |
828 | icmp_param.skb = skb; | |
829 | icmp_param.offset = 0; | |
830 | icmp_param.data_len = 0; | |
831 | icmp_param.head_len = sizeof(struct icmphdr) + 12; | |
832 | icmp_reply(&icmp_param, skb); | |
833 | out: | |
834 | return; | |
835 | out_err: | |
adf30907 | 836 | ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS); |
1da177e4 LT |
837 | goto out; |
838 | } | |
839 | ||
840 | ||
841 | /* | |
842 | * Handle ICMP_ADDRESS_MASK requests. (RFC950) | |
843 | * | |
844 | * RFC1122 (3.2.2.9). A host MUST only send replies to | |
845 | * ADDRESS_MASK requests if it's been configured as an address mask | |
846 | * agent. Receiving a request doesn't constitute implicit permission to | |
847 | * act as one. Of course, implementing this correctly requires (SHOULD) | |
848 | * a way to turn the functionality on and off. Another one for sysctl(), | |
849 | * I guess. -- MS | |
850 | * | |
851 | * RFC1812 (4.3.3.9). A router MUST implement it. | |
852 | * A router SHOULD have switch turning it on/off. | |
853 | * This switch MUST be ON by default. | |
854 | * | |
855 | * Gratuitous replies, zero-source replies are not implemented, | |
856 | * that complies with RFC. DO NOT implement them!!! All the idea | |
857 | * of broadcast addrmask replies as specified in RFC950 is broken. | |
858 | * The problem is that it is not uncommon to have several prefixes | |
859 | * on one physical interface. Moreover, addrmask agent can even be | |
860 | * not aware of existing another prefixes. | |
861 | * If source is zero, addrmask agent cannot choose correct prefix. | |
862 | * Gratuitous mask announcements suffer from the same problem. | |
863 | * RFC1812 explains it, but still allows to use ADDRMASK, | |
864 | * that is pretty silly. --ANK | |
865 | * | |
866 | * All these rules are so bizarre, that I removed kernel addrmask | |
867 | * support at all. It is wrong, it is obsolete, nobody uses it in | |
868 | * any case. --ANK | |
869 | * | |
870 | * Furthermore you can do it with a usermode address agent program | |
871 | * anyway... | |
872 | */ | |
873 | ||
874 | static void icmp_address(struct sk_buff *skb) | |
875 | { | |
876 | #if 0 | |
e87cc472 | 877 | net_dbg_ratelimited("a guy asks for address mask. Who is it?\n"); |
1da177e4 LT |
878 | #endif |
879 | } | |
880 | ||
881 | /* | |
882 | * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain | |
883 | * loudly if an inconsistency is found. | |
cfa087f6 | 884 | * called with rcu_read_lock() |
1da177e4 LT |
885 | */ |
886 | ||
887 | static void icmp_address_reply(struct sk_buff *skb) | |
888 | { | |
511c3f92 | 889 | struct rtable *rt = skb_rtable(skb); |
1da177e4 LT |
890 | struct net_device *dev = skb->dev; |
891 | struct in_device *in_dev; | |
892 | struct in_ifaddr *ifa; | |
893 | ||
894 | if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC)) | |
cfa087f6 | 895 | return; |
1da177e4 | 896 | |
cfa087f6 | 897 | in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 898 | if (!in_dev) |
cfa087f6 ED |
899 | return; |
900 | ||
1da177e4 LT |
901 | if (in_dev->ifa_list && |
902 | IN_DEV_LOG_MARTIANS(in_dev) && | |
903 | IN_DEV_FORWARD(in_dev)) { | |
a144ea4b | 904 | __be32 _mask, *mp; |
1da177e4 LT |
905 | |
906 | mp = skb_header_pointer(skb, 0, sizeof(_mask), &_mask); | |
09a62660 | 907 | BUG_ON(mp == NULL); |
1da177e4 LT |
908 | for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { |
909 | if (*mp == ifa->ifa_mask && | |
9f6abb5f | 910 | inet_ifa_match(ip_hdr(skb)->saddr, ifa)) |
1da177e4 LT |
911 | break; |
912 | } | |
e87cc472 JP |
913 | if (!ifa) |
914 | net_info_ratelimited("Wrong address mask %pI4 from %s/%pI4\n", | |
915 | mp, | |
916 | dev->name, &ip_hdr(skb)->saddr); | |
1da177e4 | 917 | } |
1da177e4 LT |
918 | } |
919 | ||
920 | static void icmp_discard(struct sk_buff *skb) | |
921 | { | |
922 | } | |
923 | ||
924 | /* | |
925 | * Deal with incoming ICMP packets. | |
926 | */ | |
927 | int icmp_rcv(struct sk_buff *skb) | |
928 | { | |
929 | struct icmphdr *icmph; | |
511c3f92 | 930 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 931 | struct net *net = dev_net(rt->dst.dev); |
1da177e4 | 932 | |
aebcf82c | 933 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { |
def8b4fa | 934 | struct sec_path *sp = skb_sec_path(skb); |
8b7817f3 HX |
935 | int nh; |
936 | ||
def8b4fa | 937 | if (!(sp && sp->xvec[sp->len - 1]->props.flags & |
aebcf82c HX |
938 | XFRM_STATE_ICMP)) |
939 | goto drop; | |
940 | ||
8b7817f3 HX |
941 | if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr))) |
942 | goto drop; | |
943 | ||
944 | nh = skb_network_offset(skb); | |
945 | skb_set_network_header(skb, sizeof(*icmph)); | |
946 | ||
947 | if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) | |
948 | goto drop; | |
949 | ||
950 | skb_set_network_header(skb, nh); | |
951 | } | |
952 | ||
dcfc23ca | 953 | ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS); |
1da177e4 LT |
954 | |
955 | switch (skb->ip_summed) { | |
84fa7933 | 956 | case CHECKSUM_COMPLETE: |
d3bc23e7 | 957 | if (!csum_fold(skb->csum)) |
1da177e4 | 958 | break; |
fb286bb2 | 959 | /* fall through */ |
1da177e4 | 960 | case CHECKSUM_NONE: |
fb286bb2 HX |
961 | skb->csum = 0; |
962 | if (__skb_checksum_complete(skb)) | |
1da177e4 | 963 | goto error; |
1da177e4 LT |
964 | } |
965 | ||
8cf22943 HX |
966 | if (!pskb_pull(skb, sizeof(*icmph))) |
967 | goto error; | |
1da177e4 | 968 | |
88c7664f | 969 | icmph = icmp_hdr(skb); |
1da177e4 | 970 | |
f66ac03d | 971 | ICMPMSGIN_INC_STATS_BH(net, icmph->type); |
1da177e4 LT |
972 | /* |
973 | * 18 is the highest 'known' ICMP type. Anything else is a mystery | |
974 | * | |
975 | * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently | |
976 | * discarded. | |
977 | */ | |
978 | if (icmph->type > NR_ICMP_TYPES) | |
979 | goto error; | |
980 | ||
981 | ||
982 | /* | |
983 | * Parse the ICMP message | |
984 | */ | |
985 | ||
e905a9ed | 986 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
1da177e4 LT |
987 | /* |
988 | * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be | |
989 | * silently ignored (we let user decide with a sysctl). | |
990 | * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently | |
991 | * discarded if to broadcast/multicast. | |
992 | */ | |
4c866aa7 AK |
993 | if ((icmph->type == ICMP_ECHO || |
994 | icmph->type == ICMP_TIMESTAMP) && | |
b34a95ee | 995 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts) { |
1da177e4 LT |
996 | goto error; |
997 | } | |
998 | if (icmph->type != ICMP_ECHO && | |
999 | icmph->type != ICMP_TIMESTAMP && | |
1000 | icmph->type != ICMP_ADDRESS && | |
1001 | icmph->type != ICMP_ADDRESSREPLY) { | |
1002 | goto error; | |
e905a9ed | 1003 | } |
1da177e4 LT |
1004 | } |
1005 | ||
1da177e4 LT |
1006 | icmp_pointers[icmph->type].handler(skb); |
1007 | ||
1008 | drop: | |
1009 | kfree_skb(skb); | |
1010 | return 0; | |
1011 | error: | |
dcfc23ca | 1012 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
1013 | goto drop; |
1014 | } | |
1015 | ||
1016 | /* | |
1017 | * This table is the definition of how we handle ICMP. | |
1018 | */ | |
9b5b5cff | 1019 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = { |
1da177e4 | 1020 | [ICMP_ECHOREPLY] = { |
c319b4d7 | 1021 | .handler = ping_rcv, |
1da177e4 LT |
1022 | }, |
1023 | [1] = { | |
1da177e4 LT |
1024 | .handler = icmp_discard, |
1025 | .error = 1, | |
1026 | }, | |
1027 | [2] = { | |
1da177e4 LT |
1028 | .handler = icmp_discard, |
1029 | .error = 1, | |
1030 | }, | |
1031 | [ICMP_DEST_UNREACH] = { | |
1da177e4 LT |
1032 | .handler = icmp_unreach, |
1033 | .error = 1, | |
1034 | }, | |
1035 | [ICMP_SOURCE_QUENCH] = { | |
1da177e4 LT |
1036 | .handler = icmp_unreach, |
1037 | .error = 1, | |
1038 | }, | |
1039 | [ICMP_REDIRECT] = { | |
1da177e4 LT |
1040 | .handler = icmp_redirect, |
1041 | .error = 1, | |
1042 | }, | |
1043 | [6] = { | |
1da177e4 LT |
1044 | .handler = icmp_discard, |
1045 | .error = 1, | |
1046 | }, | |
1047 | [7] = { | |
1da177e4 LT |
1048 | .handler = icmp_discard, |
1049 | .error = 1, | |
1050 | }, | |
1051 | [ICMP_ECHO] = { | |
1da177e4 LT |
1052 | .handler = icmp_echo, |
1053 | }, | |
1054 | [9] = { | |
1da177e4 LT |
1055 | .handler = icmp_discard, |
1056 | .error = 1, | |
1057 | }, | |
1058 | [10] = { | |
1da177e4 LT |
1059 | .handler = icmp_discard, |
1060 | .error = 1, | |
1061 | }, | |
1062 | [ICMP_TIME_EXCEEDED] = { | |
1da177e4 LT |
1063 | .handler = icmp_unreach, |
1064 | .error = 1, | |
1065 | }, | |
1066 | [ICMP_PARAMETERPROB] = { | |
1da177e4 LT |
1067 | .handler = icmp_unreach, |
1068 | .error = 1, | |
1069 | }, | |
1070 | [ICMP_TIMESTAMP] = { | |
1da177e4 LT |
1071 | .handler = icmp_timestamp, |
1072 | }, | |
1073 | [ICMP_TIMESTAMPREPLY] = { | |
1da177e4 LT |
1074 | .handler = icmp_discard, |
1075 | }, | |
1076 | [ICMP_INFO_REQUEST] = { | |
1da177e4 LT |
1077 | .handler = icmp_discard, |
1078 | }, | |
e905a9ed | 1079 | [ICMP_INFO_REPLY] = { |
1da177e4 LT |
1080 | .handler = icmp_discard, |
1081 | }, | |
1082 | [ICMP_ADDRESS] = { | |
1da177e4 LT |
1083 | .handler = icmp_address, |
1084 | }, | |
1085 | [ICMP_ADDRESSREPLY] = { | |
1da177e4 LT |
1086 | .handler = icmp_address_reply, |
1087 | }, | |
1088 | }; | |
1089 | ||
4a6ad7a1 | 1090 | static void __net_exit icmp_sk_exit(struct net *net) |
1da177e4 | 1091 | { |
1da177e4 LT |
1092 | int i; |
1093 | ||
5c8cafd6 | 1094 | for_each_possible_cpu(i) |
c1e9894d | 1095 | inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]); |
4a6ad7a1 DL |
1096 | kfree(net->ipv4.icmp_sk); |
1097 | net->ipv4.icmp_sk = NULL; | |
a5710d65 | 1098 | } |
1da177e4 | 1099 | |
263173af | 1100 | static int __net_init icmp_sk_init(struct net *net) |
a5710d65 | 1101 | { |
a5710d65 DL |
1102 | int i, err; |
1103 | ||
4a6ad7a1 DL |
1104 | net->ipv4.icmp_sk = |
1105 | kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL); | |
1106 | if (net->ipv4.icmp_sk == NULL) | |
79c91159 DL |
1107 | return -ENOMEM; |
1108 | ||
a5710d65 | 1109 | for_each_possible_cpu(i) { |
1e3cf683 | 1110 | struct sock *sk; |
1da177e4 | 1111 | |
c1e9894d DL |
1112 | err = inet_ctl_sock_create(&sk, PF_INET, |
1113 | SOCK_RAW, IPPROTO_ICMP, net); | |
1da177e4 | 1114 | if (err < 0) |
a5710d65 | 1115 | goto fail; |
1da177e4 | 1116 | |
c1e9894d | 1117 | net->ipv4.icmp_sk[i] = sk; |
1da177e4 LT |
1118 | |
1119 | /* Enough space for 2 64K ICMP packets, including | |
87fb4b7b | 1120 | * sk_buff/skb_shared_info struct overhead. |
1da177e4 | 1121 | */ |
87fb4b7b | 1122 | sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); |
1da177e4 | 1123 | |
b3a5b6cc ED |
1124 | /* |
1125 | * Speedup sock_wfree() | |
1126 | */ | |
1127 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
c1e9894d | 1128 | inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT; |
1da177e4 | 1129 | } |
a24022e1 PE |
1130 | |
1131 | /* Control parameters for ECHO replies. */ | |
1132 | net->ipv4.sysctl_icmp_echo_ignore_all = 0; | |
1133 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1; | |
1134 | ||
1135 | /* Control parameter - ignore bogus broadcast responses? */ | |
1136 | net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1; | |
1137 | ||
1138 | /* | |
1139 | * Configurable global rate limit. | |
1140 | * | |
1141 | * ratelimit defines tokens/packet consumed for dst->rate_token | |
1142 | * bucket ratemask defines which icmp types are ratelimited by | |
1143 | * setting it's bit position. | |
1144 | * | |
1145 | * default: | |
1146 | * dest unreachable (3), source quench (4), | |
1147 | * time exceeded (11), parameter problem (12) | |
1148 | */ | |
1149 | ||
1150 | net->ipv4.sysctl_icmp_ratelimit = 1 * HZ; | |
1151 | net->ipv4.sysctl_icmp_ratemask = 0x1818; | |
1152 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0; | |
1153 | ||
a5710d65 DL |
1154 | return 0; |
1155 | ||
1156 | fail: | |
1d1c8d13 | 1157 | for_each_possible_cpu(i) |
c1e9894d | 1158 | inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]); |
1d1c8d13 | 1159 | kfree(net->ipv4.icmp_sk); |
a5710d65 | 1160 | return err; |
1da177e4 LT |
1161 | } |
1162 | ||
4a6ad7a1 DL |
1163 | static struct pernet_operations __net_initdata icmp_sk_ops = { |
1164 | .init = icmp_sk_init, | |
1165 | .exit = icmp_sk_exit, | |
1166 | }; | |
1167 | ||
1168 | int __init icmp_init(void) | |
1169 | { | |
959d2726 | 1170 | return register_pernet_subsys(&icmp_sk_ops); |
4a6ad7a1 | 1171 | } |