]>
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> |
613d09b3 | 99 | #include <net/vrf.h> |
1da177e4 LT |
100 | |
101 | /* | |
102 | * Build xmit assembly blocks | |
103 | */ | |
104 | ||
105 | struct icmp_bxm { | |
106 | struct sk_buff *skb; | |
107 | int offset; | |
108 | int data_len; | |
109 | ||
110 | struct { | |
111 | struct icmphdr icmph; | |
b03d73e3 | 112 | __be32 times[3]; |
1da177e4 LT |
113 | } data; |
114 | int head_len; | |
f6d8bd05 | 115 | struct ip_options_data replyopts; |
1da177e4 LT |
116 | }; |
117 | ||
1da177e4 LT |
118 | /* An array of errno for error messages from dest unreach. */ |
119 | /* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */ | |
120 | ||
e754834e | 121 | const struct icmp_err icmp_err_convert[] = { |
1da177e4 LT |
122 | { |
123 | .errno = ENETUNREACH, /* ICMP_NET_UNREACH */ | |
124 | .fatal = 0, | |
125 | }, | |
126 | { | |
127 | .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */ | |
128 | .fatal = 0, | |
129 | }, | |
130 | { | |
131 | .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */, | |
132 | .fatal = 1, | |
133 | }, | |
134 | { | |
135 | .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */ | |
136 | .fatal = 1, | |
137 | }, | |
138 | { | |
139 | .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */ | |
140 | .fatal = 0, | |
141 | }, | |
142 | { | |
143 | .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */ | |
144 | .fatal = 0, | |
145 | }, | |
146 | { | |
147 | .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */ | |
148 | .fatal = 1, | |
149 | }, | |
150 | { | |
151 | .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */ | |
152 | .fatal = 1, | |
153 | }, | |
154 | { | |
155 | .errno = ENONET, /* ICMP_HOST_ISOLATED */ | |
156 | .fatal = 1, | |
157 | }, | |
158 | { | |
159 | .errno = ENETUNREACH, /* ICMP_NET_ANO */ | |
160 | .fatal = 1, | |
161 | }, | |
162 | { | |
163 | .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */ | |
164 | .fatal = 1, | |
165 | }, | |
166 | { | |
167 | .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */ | |
168 | .fatal = 0, | |
169 | }, | |
170 | { | |
171 | .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */ | |
172 | .fatal = 0, | |
173 | }, | |
174 | { | |
175 | .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */ | |
176 | .fatal = 1, | |
177 | }, | |
178 | { | |
179 | .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */ | |
180 | .fatal = 1, | |
181 | }, | |
182 | { | |
183 | .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */ | |
184 | .fatal = 1, | |
185 | }, | |
186 | }; | |
4bc2f18b | 187 | EXPORT_SYMBOL(icmp_err_convert); |
1da177e4 | 188 | |
1da177e4 LT |
189 | /* |
190 | * ICMP control array. This specifies what to do with each ICMP. | |
191 | */ | |
192 | ||
193 | struct icmp_control { | |
e3e32170 | 194 | bool (*handler)(struct sk_buff *skb); |
1da177e4 LT |
195 | short error; /* This ICMP is classed as an error message */ |
196 | }; | |
197 | ||
9b5b5cff | 198 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1]; |
1da177e4 LT |
199 | |
200 | /* | |
201 | * The ICMP socket(s). This is the most convenient way to flow control | |
202 | * our ICMP output as well as maintain a clean interface throughout | |
203 | * all layers. All Socketless IP sends will soon be gone. | |
204 | * | |
205 | * On SMP we have one ICMP socket per-cpu. | |
206 | */ | |
4a6ad7a1 DL |
207 | static struct sock *icmp_sk(struct net *net) |
208 | { | |
349c9e3c | 209 | return *this_cpu_ptr(net->ipv4.icmp_sk); |
4a6ad7a1 | 210 | } |
1da177e4 | 211 | |
fdc0bde9 | 212 | static inline struct sock *icmp_xmit_lock(struct net *net) |
1da177e4 | 213 | { |
fdc0bde9 DL |
214 | struct sock *sk; |
215 | ||
1da177e4 LT |
216 | local_bh_disable(); |
217 | ||
fdc0bde9 DL |
218 | sk = icmp_sk(net); |
219 | ||
405666db | 220 | if (unlikely(!spin_trylock(&sk->sk_lock.slock))) { |
1da177e4 LT |
221 | /* This can happen if the output path signals a |
222 | * dst_link_failure() for an outgoing ICMP packet. | |
223 | */ | |
224 | local_bh_enable(); | |
fdc0bde9 | 225 | return NULL; |
1da177e4 | 226 | } |
fdc0bde9 | 227 | return sk; |
1da177e4 LT |
228 | } |
229 | ||
405666db | 230 | static inline void icmp_xmit_unlock(struct sock *sk) |
1da177e4 | 231 | { |
405666db | 232 | spin_unlock_bh(&sk->sk_lock.slock); |
1da177e4 LT |
233 | } |
234 | ||
4cdf507d ED |
235 | int sysctl_icmp_msgs_per_sec __read_mostly = 1000; |
236 | int sysctl_icmp_msgs_burst __read_mostly = 50; | |
237 | ||
238 | static struct { | |
239 | spinlock_t lock; | |
240 | u32 credit; | |
241 | u32 stamp; | |
242 | } icmp_global = { | |
243 | .lock = __SPIN_LOCK_UNLOCKED(icmp_global.lock), | |
244 | }; | |
245 | ||
246 | /** | |
247 | * icmp_global_allow - Are we allowed to send one more ICMP message ? | |
248 | * | |
249 | * Uses a token bucket to limit our ICMP messages to sysctl_icmp_msgs_per_sec. | |
250 | * Returns false if we reached the limit and can not send another packet. | |
251 | * Note: called with BH disabled | |
252 | */ | |
253 | bool icmp_global_allow(void) | |
254 | { | |
255 | u32 credit, delta, incr = 0, now = (u32)jiffies; | |
256 | bool rc = false; | |
257 | ||
258 | /* Check if token bucket is empty and cannot be refilled | |
259 | * without taking the spinlock. | |
260 | */ | |
261 | if (!icmp_global.credit) { | |
262 | delta = min_t(u32, now - icmp_global.stamp, HZ); | |
263 | if (delta < HZ / 50) | |
264 | return false; | |
265 | } | |
266 | ||
267 | spin_lock(&icmp_global.lock); | |
268 | delta = min_t(u32, now - icmp_global.stamp, HZ); | |
269 | if (delta >= HZ / 50) { | |
270 | incr = sysctl_icmp_msgs_per_sec * delta / HZ ; | |
271 | if (incr) | |
272 | icmp_global.stamp = now; | |
273 | } | |
274 | credit = min_t(u32, icmp_global.credit + incr, sysctl_icmp_msgs_burst); | |
275 | if (credit) { | |
276 | credit--; | |
277 | rc = true; | |
278 | } | |
279 | icmp_global.credit = credit; | |
280 | spin_unlock(&icmp_global.lock); | |
281 | return rc; | |
282 | } | |
283 | EXPORT_SYMBOL(icmp_global_allow); | |
284 | ||
1da177e4 LT |
285 | /* |
286 | * Send an ICMP frame. | |
287 | */ | |
288 | ||
4cdf507d ED |
289 | static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt, |
290 | struct flowi4 *fl4, int type, int code) | |
1da177e4 | 291 | { |
d8d1f30b | 292 | struct dst_entry *dst = &rt->dst; |
92d86829 | 293 | bool rc = true; |
1da177e4 LT |
294 | |
295 | if (type > NR_ICMP_TYPES) | |
296 | goto out; | |
297 | ||
298 | /* Don't limit PMTU discovery. */ | |
299 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
300 | goto out; | |
301 | ||
302 | /* No rate limit on loopback */ | |
303 | if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) | |
e905a9ed | 304 | goto out; |
1da177e4 LT |
305 | |
306 | /* Limit if icmp type is enabled in ratemask. */ | |
4cdf507d ED |
307 | if (!((1 << type) & net->ipv4.sysctl_icmp_ratemask)) |
308 | goto out; | |
309 | ||
310 | rc = false; | |
311 | if (icmp_global_allow()) { | |
312 | struct inet_peer *peer; | |
313 | ||
314 | peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, 1); | |
fbfe95a4 | 315 | rc = inet_peer_xrlim_allow(peer, |
92d86829 | 316 | net->ipv4.sysctl_icmp_ratelimit); |
e1a67642 NC |
317 | if (peer) |
318 | inet_putpeer(peer); | |
92d86829 | 319 | } |
1da177e4 LT |
320 | out: |
321 | return rc; | |
322 | } | |
323 | ||
324 | /* | |
325 | * Maintain the counters used in the SNMP statistics for outgoing ICMP | |
326 | */ | |
0388b004 | 327 | void icmp_out_count(struct net *net, unsigned char type) |
1da177e4 | 328 | { |
903fc196 | 329 | ICMPMSGOUT_INC_STATS(net, type); |
75c939bb | 330 | ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS); |
1da177e4 LT |
331 | } |
332 | ||
333 | /* | |
334 | * Checksum each fragment, and on the first include the headers and final | |
335 | * checksum. | |
336 | */ | |
337 | static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd, | |
338 | struct sk_buff *skb) | |
339 | { | |
340 | struct icmp_bxm *icmp_param = (struct icmp_bxm *)from; | |
5f92a738 | 341 | __wsum csum; |
1da177e4 LT |
342 | |
343 | csum = skb_copy_and_csum_bits(icmp_param->skb, | |
344 | icmp_param->offset + offset, | |
345 | to, len, 0); | |
346 | ||
347 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
348 | if (icmp_pointers[icmp_param->data.icmph.type].error) | |
349 | nf_ct_attach(skb, icmp_param->skb); | |
350 | return 0; | |
351 | } | |
352 | ||
353 | static void icmp_push_reply(struct icmp_bxm *icmp_param, | |
77968b78 | 354 | struct flowi4 *fl4, |
2e77d89b | 355 | struct ipcm_cookie *ipc, struct rtable **rt) |
1da177e4 | 356 | { |
1e3cf683 | 357 | struct sock *sk; |
1da177e4 LT |
358 | struct sk_buff *skb; |
359 | ||
d8d1f30b | 360 | sk = icmp_sk(dev_net((*rt)->dst.dev)); |
f5fca608 | 361 | if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param, |
e905a9ed YH |
362 | icmp_param->data_len+icmp_param->head_len, |
363 | icmp_param->head_len, | |
1f8438a8 ED |
364 | ipc, rt, MSG_DONTWAIT) < 0) { |
365 | ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS); | |
1e3cf683 | 366 | ip_flush_pending_frames(sk); |
1f8438a8 | 367 | } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) { |
88c7664f | 368 | struct icmphdr *icmph = icmp_hdr(skb); |
d3bc23e7 | 369 | __wsum csum = 0; |
1da177e4 LT |
370 | struct sk_buff *skb1; |
371 | ||
1e3cf683 | 372 | skb_queue_walk(&sk->sk_write_queue, skb1) { |
1da177e4 LT |
373 | csum = csum_add(csum, skb1->csum); |
374 | } | |
375 | csum = csum_partial_copy_nocheck((void *)&icmp_param->data, | |
376 | (char *)icmph, | |
377 | icmp_param->head_len, csum); | |
378 | icmph->checksum = csum_fold(csum); | |
379 | skb->ip_summed = CHECKSUM_NONE; | |
77968b78 | 380 | ip_push_pending_frames(sk, fl4); |
1da177e4 LT |
381 | } |
382 | } | |
383 | ||
384 | /* | |
385 | * Driving logic for building and sending ICMP messages. | |
386 | */ | |
387 | ||
388 | static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb) | |
389 | { | |
1da177e4 | 390 | struct ipcm_cookie ipc; |
511c3f92 | 391 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 392 | struct net *net = dev_net(rt->dst.dev); |
77968b78 | 393 | struct flowi4 fl4; |
fdc0bde9 DL |
394 | struct sock *sk; |
395 | struct inet_sock *inet; | |
35ebf65e | 396 | __be32 daddr, saddr; |
e110861f | 397 | u32 mark = IP4_REPLY_MARK(net, skb->mark); |
1da177e4 | 398 | |
f6d8bd05 | 399 | if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb)) |
f00c401b | 400 | return; |
1da177e4 | 401 | |
fdc0bde9 | 402 | sk = icmp_xmit_lock(net); |
51456b29 | 403 | if (!sk) |
1da177e4 | 404 | return; |
fdc0bde9 | 405 | inet = inet_sk(sk); |
1da177e4 LT |
406 | |
407 | icmp_param->data.icmph.checksum = 0; | |
1da177e4 | 408 | |
eddc9ec5 | 409 | inet->tos = ip_hdr(skb)->tos; |
e110861f | 410 | sk->sk_mark = mark; |
9f6abb5f | 411 | daddr = ipc.addr = ip_hdr(skb)->saddr; |
35ebf65e | 412 | saddr = fib_compute_spec_dst(skb); |
1da177e4 | 413 | ipc.opt = NULL; |
2244d07b | 414 | ipc.tx_flags = 0; |
aa661581 FF |
415 | ipc.ttl = 0; |
416 | ipc.tos = -1; | |
417 | ||
f6d8bd05 ED |
418 | if (icmp_param->replyopts.opt.opt.optlen) { |
419 | ipc.opt = &icmp_param->replyopts.opt; | |
420 | if (ipc.opt->opt.srr) | |
421 | daddr = icmp_param->replyopts.opt.opt.faddr; | |
1da177e4 | 422 | } |
77968b78 DM |
423 | memset(&fl4, 0, sizeof(fl4)); |
424 | fl4.daddr = daddr; | |
35ebf65e | 425 | fl4.saddr = saddr; |
e110861f | 426 | fl4.flowi4_mark = mark; |
77968b78 DM |
427 | fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos); |
428 | fl4.flowi4_proto = IPPROTO_ICMP; | |
613d09b3 | 429 | fl4.flowi4_oif = vrf_master_ifindex_rcu(skb->dev) ? : skb->dev->ifindex; |
77968b78 DM |
430 | security_skb_classify_flow(skb, flowi4_to_flowi(&fl4)); |
431 | rt = ip_route_output_key(net, &fl4); | |
432 | if (IS_ERR(rt)) | |
433 | goto out_unlock; | |
a48eff12 | 434 | if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type, |
1da177e4 | 435 | icmp_param->data.icmph.code)) |
77968b78 | 436 | icmp_push_reply(icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
437 | ip_rt_put(rt); |
438 | out_unlock: | |
405666db | 439 | icmp_xmit_unlock(sk); |
1da177e4 LT |
440 | } |
441 | ||
77968b78 DM |
442 | static struct rtable *icmp_route_lookup(struct net *net, |
443 | struct flowi4 *fl4, | |
444 | struct sk_buff *skb_in, | |
b71d1d42 | 445 | const struct iphdr *iph, |
e110861f | 446 | __be32 saddr, u8 tos, u32 mark, |
f6d460cf DM |
447 | int type, int code, |
448 | struct icmp_bxm *param) | |
449 | { | |
f6d460cf | 450 | struct rtable *rt, *rt2; |
415b3334 | 451 | struct flowi4 fl4_dec; |
f6d460cf DM |
452 | int err; |
453 | ||
77968b78 DM |
454 | memset(fl4, 0, sizeof(*fl4)); |
455 | fl4->daddr = (param->replyopts.opt.opt.srr ? | |
456 | param->replyopts.opt.opt.faddr : iph->saddr); | |
457 | fl4->saddr = saddr; | |
e110861f | 458 | fl4->flowi4_mark = mark; |
77968b78 DM |
459 | fl4->flowi4_tos = RT_TOS(tos); |
460 | fl4->flowi4_proto = IPPROTO_ICMP; | |
461 | fl4->fl4_icmp_type = type; | |
462 | fl4->fl4_icmp_code = code; | |
613d09b3 DA |
463 | fl4->flowi4_oif = vrf_master_ifindex_rcu(skb_in->dev) ? : skb_in->dev->ifindex; |
464 | ||
77968b78 DM |
465 | security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4)); |
466 | rt = __ip_route_output_key(net, fl4); | |
b23dd4fe DM |
467 | if (IS_ERR(rt)) |
468 | return rt; | |
f6d460cf DM |
469 | |
470 | /* No need to clone since we're just using its address. */ | |
471 | rt2 = rt; | |
472 | ||
9d6ec938 | 473 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
77968b78 | 474 | flowi4_to_flowi(fl4), NULL, 0); |
452edd59 | 475 | if (!IS_ERR(rt)) { |
f6d460cf DM |
476 | if (rt != rt2) |
477 | return rt; | |
452edd59 | 478 | } else if (PTR_ERR(rt) == -EPERM) { |
f6d460cf | 479 | rt = NULL; |
452edd59 DM |
480 | } else |
481 | return rt; | |
f6d460cf | 482 | |
415b3334 | 483 | err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET); |
f6d460cf DM |
484 | if (err) |
485 | goto relookup_failed; | |
486 | ||
415b3334 DM |
487 | if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) { |
488 | rt2 = __ip_route_output_key(net, &fl4_dec); | |
b23dd4fe DM |
489 | if (IS_ERR(rt2)) |
490 | err = PTR_ERR(rt2); | |
f6d460cf | 491 | } else { |
9d6ec938 | 492 | struct flowi4 fl4_2 = {}; |
f6d460cf DM |
493 | unsigned long orefdst; |
494 | ||
415b3334 | 495 | fl4_2.daddr = fl4_dec.saddr; |
9d6ec938 | 496 | rt2 = ip_route_output_key(net, &fl4_2); |
b23dd4fe DM |
497 | if (IS_ERR(rt2)) { |
498 | err = PTR_ERR(rt2); | |
f6d460cf | 499 | goto relookup_failed; |
b23dd4fe | 500 | } |
f6d460cf DM |
501 | /* Ugh! */ |
502 | orefdst = skb_in->_skb_refdst; /* save old refdst */ | |
773a69d6 | 503 | skb_dst_set(skb_in, NULL); |
415b3334 | 504 | err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr, |
f6d460cf DM |
505 | RT_TOS(tos), rt2->dst.dev); |
506 | ||
507 | dst_release(&rt2->dst); | |
508 | rt2 = skb_rtable(skb_in); | |
509 | skb_in->_skb_refdst = orefdst; /* restore old refdst */ | |
510 | } | |
511 | ||
512 | if (err) | |
513 | goto relookup_failed; | |
514 | ||
9d6ec938 | 515 | rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst, |
415b3334 | 516 | flowi4_to_flowi(&fl4_dec), NULL, |
9d6ec938 | 517 | XFRM_LOOKUP_ICMP); |
452edd59 | 518 | if (!IS_ERR(rt2)) { |
f6d460cf | 519 | dst_release(&rt->dst); |
415b3334 | 520 | memcpy(fl4, &fl4_dec, sizeof(*fl4)); |
f6d460cf | 521 | rt = rt2; |
452edd59 DM |
522 | } else if (PTR_ERR(rt2) == -EPERM) { |
523 | if (rt) | |
524 | dst_release(&rt->dst); | |
525 | return rt2; | |
526 | } else { | |
527 | err = PTR_ERR(rt2); | |
528 | goto relookup_failed; | |
f6d460cf | 529 | } |
f6d460cf DM |
530 | return rt; |
531 | ||
532 | relookup_failed: | |
533 | if (rt) | |
534 | return rt; | |
535 | return ERR_PTR(err); | |
536 | } | |
1da177e4 LT |
537 | |
538 | /* | |
539 | * Send an ICMP message in response to a situation | |
540 | * | |
541 | * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header. | |
542 | * MAY send more (we do). | |
543 | * MUST NOT change this header information. | |
544 | * MUST NOT reply to a multicast/broadcast IP address. | |
545 | * MUST NOT reply to a multicast/broadcast MAC address. | |
546 | * MUST reply to only the first fragment. | |
547 | */ | |
548 | ||
e4883014 | 549 | void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info) |
1da177e4 LT |
550 | { |
551 | struct iphdr *iph; | |
552 | int room; | |
9a99d4a5 | 553 | struct icmp_bxm *icmp_param; |
511c3f92 | 554 | struct rtable *rt = skb_rtable(skb_in); |
1da177e4 | 555 | struct ipcm_cookie ipc; |
77968b78 | 556 | struct flowi4 fl4; |
a61ced5d | 557 | __be32 saddr; |
1da177e4 | 558 | u8 tos; |
e110861f | 559 | u32 mark; |
dde1bc0e | 560 | struct net *net; |
4a6ad7a1 | 561 | struct sock *sk; |
1da177e4 LT |
562 | |
563 | if (!rt) | |
564 | goto out; | |
d8d1f30b | 565 | net = dev_net(rt->dst.dev); |
1da177e4 LT |
566 | |
567 | /* | |
568 | * Find the original header. It is expected to be valid, of course. | |
569 | * Check this, icmp_send is called from the most obscure devices | |
570 | * sometimes. | |
571 | */ | |
eddc9ec5 | 572 | iph = ip_hdr(skb_in); |
1da177e4 | 573 | |
27a884dc | 574 | if ((u8 *)iph < skb_in->head || |
f7c0c2ae SH |
575 | (skb_network_header(skb_in) + sizeof(*iph)) > |
576 | skb_tail_pointer(skb_in)) | |
1da177e4 LT |
577 | goto out; |
578 | ||
579 | /* | |
580 | * No replies to physical multicast/broadcast | |
581 | */ | |
582 | if (skb_in->pkt_type != PACKET_HOST) | |
583 | goto out; | |
584 | ||
585 | /* | |
586 | * Now check at the protocol level | |
587 | */ | |
588 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
589 | goto out; | |
590 | ||
591 | /* | |
592 | * Only reply to fragment 0. We byte re-order the constant | |
593 | * mask for efficiency. | |
594 | */ | |
595 | if (iph->frag_off & htons(IP_OFFSET)) | |
596 | goto out; | |
597 | ||
598 | /* | |
599 | * If we send an ICMP error to an ICMP error a mess would result.. | |
600 | */ | |
601 | if (icmp_pointers[type].error) { | |
602 | /* | |
603 | * We are an error, check if we are replying to an | |
604 | * ICMP error | |
605 | */ | |
606 | if (iph->protocol == IPPROTO_ICMP) { | |
607 | u8 _inner_type, *itp; | |
608 | ||
609 | itp = skb_header_pointer(skb_in, | |
d56f90a7 | 610 | skb_network_header(skb_in) + |
1da177e4 LT |
611 | (iph->ihl << 2) + |
612 | offsetof(struct icmphdr, | |
613 | type) - | |
614 | skb_in->data, | |
615 | sizeof(_inner_type), | |
616 | &_inner_type); | |
51456b29 | 617 | if (!itp) |
1da177e4 LT |
618 | goto out; |
619 | ||
620 | /* | |
621 | * Assume any unknown ICMP type is an error. This | |
622 | * isn't specified by the RFC, but think about it.. | |
623 | */ | |
624 | if (*itp > NR_ICMP_TYPES || | |
625 | icmp_pointers[*itp].error) | |
626 | goto out; | |
627 | } | |
628 | } | |
629 | ||
9a99d4a5 CW |
630 | icmp_param = kmalloc(sizeof(*icmp_param), GFP_ATOMIC); |
631 | if (!icmp_param) | |
632 | return; | |
633 | ||
fdc0bde9 | 634 | sk = icmp_xmit_lock(net); |
51456b29 | 635 | if (!sk) |
9a99d4a5 | 636 | goto out_free; |
1da177e4 LT |
637 | |
638 | /* | |
639 | * Construct source address and options. | |
640 | */ | |
641 | ||
642 | saddr = iph->daddr; | |
1c2fb7f9 | 643 | if (!(rt->rt_flags & RTCF_LOCAL)) { |
6e1d9103 PM |
644 | struct net_device *dev = NULL; |
645 | ||
685c7944 | 646 | rcu_read_lock(); |
c7537967 DM |
647 | if (rt_is_input_route(rt) && |
648 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr) | |
92101b3b | 649 | dev = dev_get_by_index_rcu(net, inet_iif(skb_in)); |
6e1d9103 | 650 | |
685c7944 | 651 | if (dev) |
6e1d9103 | 652 | saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
685c7944 | 653 | else |
1c2fb7f9 | 654 | saddr = 0; |
685c7944 | 655 | rcu_read_unlock(); |
1c2fb7f9 | 656 | } |
1da177e4 LT |
657 | |
658 | tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) | | |
659 | IPTOS_PREC_INTERNETCONTROL) : | |
660 | iph->tos; | |
e110861f | 661 | mark = IP4_REPLY_MARK(net, skb_in->mark); |
1da177e4 | 662 | |
9a99d4a5 | 663 | if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb_in)) |
fa60cf7f | 664 | goto out_unlock; |
1da177e4 LT |
665 | |
666 | ||
667 | /* | |
668 | * Prepare data for ICMP header. | |
669 | */ | |
670 | ||
9a99d4a5 CW |
671 | icmp_param->data.icmph.type = type; |
672 | icmp_param->data.icmph.code = code; | |
673 | icmp_param->data.icmph.un.gateway = info; | |
674 | icmp_param->data.icmph.checksum = 0; | |
675 | icmp_param->skb = skb_in; | |
676 | icmp_param->offset = skb_network_offset(skb_in); | |
405666db | 677 | inet_sk(sk)->tos = tos; |
e110861f | 678 | sk->sk_mark = mark; |
1da177e4 | 679 | ipc.addr = iph->saddr; |
9a99d4a5 | 680 | ipc.opt = &icmp_param->replyopts.opt; |
2244d07b | 681 | ipc.tx_flags = 0; |
aa661581 FF |
682 | ipc.ttl = 0; |
683 | ipc.tos = -1; | |
1da177e4 | 684 | |
e110861f | 685 | rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, mark, |
9a99d4a5 | 686 | type, code, icmp_param); |
f6d460cf DM |
687 | if (IS_ERR(rt)) |
688 | goto out_unlock; | |
1da177e4 | 689 | |
a48eff12 | 690 | if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
1da177e4 LT |
691 | goto ende; |
692 | ||
693 | /* RFC says return as much as we can without exceeding 576 bytes. */ | |
694 | ||
d8d1f30b | 695 | room = dst_mtu(&rt->dst); |
1da177e4 LT |
696 | if (room > 576) |
697 | room = 576; | |
9a99d4a5 | 698 | room -= sizeof(struct iphdr) + icmp_param->replyopts.opt.opt.optlen; |
1da177e4 LT |
699 | room -= sizeof(struct icmphdr); |
700 | ||
9a99d4a5 CW |
701 | icmp_param->data_len = skb_in->len - icmp_param->offset; |
702 | if (icmp_param->data_len > room) | |
703 | icmp_param->data_len = room; | |
704 | icmp_param->head_len = sizeof(struct icmphdr); | |
1da177e4 | 705 | |
9a99d4a5 | 706 | icmp_push_reply(icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
707 | ende: |
708 | ip_rt_put(rt); | |
709 | out_unlock: | |
405666db | 710 | icmp_xmit_unlock(sk); |
9a99d4a5 CW |
711 | out_free: |
712 | kfree(icmp_param); | |
1da177e4 LT |
713 | out:; |
714 | } | |
4bc2f18b | 715 | EXPORT_SYMBOL(icmp_send); |
1da177e4 LT |
716 | |
717 | ||
1de9243b DM |
718 | static void icmp_socket_deliver(struct sk_buff *skb, u32 info) |
719 | { | |
720 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
721 | const struct net_protocol *ipprot; | |
722 | int protocol = iph->protocol; | |
723 | ||
f0a70e90 DM |
724 | /* Checkin full IP header plus 8 bytes of protocol to |
725 | * avoid additional coding at protocol handlers. | |
726 | */ | |
7304fe46 DJ |
727 | if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) { |
728 | ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS); | |
f0a70e90 | 729 | return; |
7304fe46 | 730 | } |
f0a70e90 | 731 | |
1de9243b DM |
732 | raw_icmp_error(skb, protocol, info); |
733 | ||
1de9243b DM |
734 | ipprot = rcu_dereference(inet_protos[protocol]); |
735 | if (ipprot && ipprot->err_handler) | |
736 | ipprot->err_handler(skb, info); | |
1de9243b DM |
737 | } |
738 | ||
8ed1dc44 HFS |
739 | static bool icmp_tag_validation(int proto) |
740 | { | |
741 | bool ok; | |
742 | ||
743 | rcu_read_lock(); | |
744 | ok = rcu_dereference(inet_protos[proto])->icmp_strict_tag_validation; | |
745 | rcu_read_unlock(); | |
746 | return ok; | |
747 | } | |
748 | ||
1da177e4 | 749 | /* |
08578d8d RR |
750 | * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, ICMP_QUENCH, and |
751 | * ICMP_PARAMETERPROB. | |
1da177e4 LT |
752 | */ |
753 | ||
e3e32170 | 754 | static bool icmp_unreach(struct sk_buff *skb) |
1da177e4 | 755 | { |
b71d1d42 | 756 | const struct iphdr *iph; |
1da177e4 | 757 | struct icmphdr *icmph; |
dde1bc0e | 758 | struct net *net; |
f9242b6b | 759 | u32 info = 0; |
dde1bc0e | 760 | |
adf30907 | 761 | net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
762 | |
763 | /* | |
764 | * Incomplete header ? | |
765 | * Only checks for the IP header, there should be an | |
766 | * additional check for longer headers in upper levels. | |
767 | */ | |
768 | ||
769 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
770 | goto out_err; | |
771 | ||
88c7664f | 772 | icmph = icmp_hdr(skb); |
b71d1d42 | 773 | iph = (const struct iphdr *)skb->data; |
1da177e4 LT |
774 | |
775 | if (iph->ihl < 5) /* Mangled header, drop. */ | |
776 | goto out_err; | |
777 | ||
778 | if (icmph->type == ICMP_DEST_UNREACH) { | |
779 | switch (icmph->code & 15) { | |
780 | case ICMP_NET_UNREACH: | |
781 | case ICMP_HOST_UNREACH: | |
782 | case ICMP_PROT_UNREACH: | |
783 | case ICMP_PORT_UNREACH: | |
784 | break; | |
785 | case ICMP_FRAG_NEEDED: | |
8ed1dc44 HFS |
786 | /* for documentation of the ip_no_pmtu_disc |
787 | * values please see | |
788 | * Documentation/networking/ip-sysctl.txt | |
789 | */ | |
790 | switch (net->ipv4.sysctl_ip_no_pmtu_disc) { | |
791 | default: | |
ba7a46f1 JP |
792 | net_dbg_ratelimited("%pI4: fragmentation needed and DF set\n", |
793 | &iph->daddr); | |
8ed1dc44 HFS |
794 | break; |
795 | case 2: | |
796 | goto out; | |
797 | case 3: | |
798 | if (!icmp_tag_validation(iph->protocol)) | |
799 | goto out; | |
800 | /* fall through */ | |
801 | case 0: | |
46517008 | 802 | info = ntohs(icmph->un.frag.mtu); |
1da177e4 LT |
803 | } |
804 | break; | |
805 | case ICMP_SR_FAILED: | |
ba7a46f1 JP |
806 | net_dbg_ratelimited("%pI4: Source Route Failed\n", |
807 | &iph->daddr); | |
1da177e4 LT |
808 | break; |
809 | default: | |
810 | break; | |
811 | } | |
812 | if (icmph->code > NR_ICMP_UNREACH) | |
813 | goto out; | |
814 | } else if (icmph->type == ICMP_PARAMETERPROB) | |
815 | info = ntohl(icmph->un.gateway) >> 24; | |
816 | ||
817 | /* | |
818 | * Throw it at our lower layers | |
819 | * | |
820 | * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed | |
821 | * header. | |
822 | * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the | |
823 | * transport layer. | |
824 | * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to | |
825 | * transport layer. | |
826 | */ | |
827 | ||
828 | /* | |
25985edc | 829 | * Check the other end isn't violating RFC 1122. Some routers send |
1da177e4 LT |
830 | * bogus responses to broadcast frames. If you see this message |
831 | * first check your netmask matches at both ends, if it does then | |
832 | * get the other vendor to fix their kit. | |
833 | */ | |
834 | ||
b34a95ee | 835 | if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses && |
dde1bc0e | 836 | inet_addr_type(net, iph->daddr) == RTN_BROADCAST) { |
e87cc472 JP |
837 | net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n", |
838 | &ip_hdr(skb)->saddr, | |
839 | icmph->type, icmph->code, | |
840 | &iph->daddr, skb->dev->name); | |
1da177e4 LT |
841 | goto out; |
842 | } | |
843 | ||
1de9243b | 844 | icmp_socket_deliver(skb, info); |
1da177e4 LT |
845 | |
846 | out: | |
e3e32170 | 847 | return true; |
1da177e4 | 848 | out_err: |
dcfc23ca | 849 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
e3e32170 | 850 | return false; |
1da177e4 LT |
851 | } |
852 | ||
853 | ||
854 | /* | |
855 | * Handle ICMP_REDIRECT. | |
856 | */ | |
857 | ||
e3e32170 | 858 | static bool icmp_redirect(struct sk_buff *skb) |
1da177e4 | 859 | { |
94206125 DM |
860 | if (skb->len < sizeof(struct iphdr)) { |
861 | ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS); | |
e3e32170 | 862 | return false; |
94206125 | 863 | } |
1da177e4 | 864 | |
e3e32170 RJ |
865 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) { |
866 | /* there aught to be a stat */ | |
867 | return false; | |
868 | } | |
c319b4d7 | 869 | |
d3351b75 | 870 | icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway); |
e3e32170 | 871 | return true; |
1da177e4 LT |
872 | } |
873 | ||
874 | /* | |
875 | * Handle ICMP_ECHO ("ping") requests. | |
876 | * | |
877 | * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo | |
878 | * requests. | |
879 | * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be | |
880 | * included in the reply. | |
881 | * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring | |
882 | * echo requests, MUST have default=NOT. | |
883 | * See also WRT handling of options once they are done and working. | |
884 | */ | |
885 | ||
e3e32170 | 886 | static bool icmp_echo(struct sk_buff *skb) |
1da177e4 | 887 | { |
b34a95ee PE |
888 | struct net *net; |
889 | ||
adf30907 | 890 | net = dev_net(skb_dst(skb)->dev); |
b34a95ee | 891 | if (!net->ipv4.sysctl_icmp_echo_ignore_all) { |
1da177e4 LT |
892 | struct icmp_bxm icmp_param; |
893 | ||
88c7664f | 894 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
895 | icmp_param.data.icmph.type = ICMP_ECHOREPLY; |
896 | icmp_param.skb = skb; | |
897 | icmp_param.offset = 0; | |
898 | icmp_param.data_len = skb->len; | |
899 | icmp_param.head_len = sizeof(struct icmphdr); | |
900 | icmp_reply(&icmp_param, skb); | |
901 | } | |
e3e32170 RJ |
902 | /* should there be an ICMP stat for ignored echos? */ |
903 | return true; | |
1da177e4 LT |
904 | } |
905 | ||
906 | /* | |
907 | * Handle ICMP Timestamp requests. | |
908 | * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests. | |
909 | * SHOULD be in the kernel for minimum random latency. | |
910 | * MUST be accurate to a few minutes. | |
911 | * MUST be updated at least at 15Hz. | |
912 | */ | |
e3e32170 | 913 | static bool icmp_timestamp(struct sk_buff *skb) |
1da177e4 | 914 | { |
f25c3d61 | 915 | struct timespec tv; |
1da177e4 LT |
916 | struct icmp_bxm icmp_param; |
917 | /* | |
918 | * Too short. | |
919 | */ | |
920 | if (skb->len < 4) | |
921 | goto out_err; | |
922 | ||
923 | /* | |
924 | * Fill in the current time as ms since midnight UT: | |
925 | */ | |
f25c3d61 YH |
926 | getnstimeofday(&tv); |
927 | icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + | |
928 | tv.tv_nsec / NSEC_PER_MSEC); | |
1da177e4 LT |
929 | icmp_param.data.times[2] = icmp_param.data.times[1]; |
930 | if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4)) | |
931 | BUG(); | |
88c7664f | 932 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
933 | icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY; |
934 | icmp_param.data.icmph.code = 0; | |
935 | icmp_param.skb = skb; | |
936 | icmp_param.offset = 0; | |
937 | icmp_param.data_len = 0; | |
938 | icmp_param.head_len = sizeof(struct icmphdr) + 12; | |
939 | icmp_reply(&icmp_param, skb); | |
e3e32170 RJ |
940 | return true; |
941 | ||
1da177e4 | 942 | out_err: |
adf30907 | 943 | ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS); |
e3e32170 | 944 | return false; |
1da177e4 LT |
945 | } |
946 | ||
e3e32170 | 947 | static bool icmp_discard(struct sk_buff *skb) |
1da177e4 | 948 | { |
e3e32170 RJ |
949 | /* pretend it was a success */ |
950 | return true; | |
1da177e4 LT |
951 | } |
952 | ||
953 | /* | |
954 | * Deal with incoming ICMP packets. | |
955 | */ | |
956 | int icmp_rcv(struct sk_buff *skb) | |
957 | { | |
958 | struct icmphdr *icmph; | |
511c3f92 | 959 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 960 | struct net *net = dev_net(rt->dst.dev); |
e3e32170 | 961 | bool success; |
1da177e4 | 962 | |
aebcf82c | 963 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { |
def8b4fa | 964 | struct sec_path *sp = skb_sec_path(skb); |
8b7817f3 HX |
965 | int nh; |
966 | ||
def8b4fa | 967 | if (!(sp && sp->xvec[sp->len - 1]->props.flags & |
aebcf82c HX |
968 | XFRM_STATE_ICMP)) |
969 | goto drop; | |
970 | ||
8b7817f3 HX |
971 | if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr))) |
972 | goto drop; | |
973 | ||
974 | nh = skb_network_offset(skb); | |
975 | skb_set_network_header(skb, sizeof(*icmph)); | |
976 | ||
977 | if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) | |
978 | goto drop; | |
979 | ||
980 | skb_set_network_header(skb, nh); | |
981 | } | |
982 | ||
dcfc23ca | 983 | ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS); |
1da177e4 | 984 | |
29a96e1f TH |
985 | if (skb_checksum_simple_validate(skb)) |
986 | goto csum_error; | |
1da177e4 | 987 | |
8cf22943 HX |
988 | if (!pskb_pull(skb, sizeof(*icmph))) |
989 | goto error; | |
1da177e4 | 990 | |
88c7664f | 991 | icmph = icmp_hdr(skb); |
1da177e4 | 992 | |
f66ac03d | 993 | ICMPMSGIN_INC_STATS_BH(net, icmph->type); |
1da177e4 LT |
994 | /* |
995 | * 18 is the highest 'known' ICMP type. Anything else is a mystery | |
996 | * | |
997 | * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently | |
998 | * discarded. | |
999 | */ | |
1000 | if (icmph->type > NR_ICMP_TYPES) | |
1001 | goto error; | |
1002 | ||
1003 | ||
1004 | /* | |
1005 | * Parse the ICMP message | |
1006 | */ | |
1007 | ||
e905a9ed | 1008 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
1da177e4 LT |
1009 | /* |
1010 | * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be | |
1011 | * silently ignored (we let user decide with a sysctl). | |
1012 | * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently | |
1013 | * discarded if to broadcast/multicast. | |
1014 | */ | |
4c866aa7 AK |
1015 | if ((icmph->type == ICMP_ECHO || |
1016 | icmph->type == ICMP_TIMESTAMP) && | |
b34a95ee | 1017 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts) { |
1da177e4 LT |
1018 | goto error; |
1019 | } | |
1020 | if (icmph->type != ICMP_ECHO && | |
1021 | icmph->type != ICMP_TIMESTAMP && | |
1022 | icmph->type != ICMP_ADDRESS && | |
1023 | icmph->type != ICMP_ADDRESSREPLY) { | |
1024 | goto error; | |
e905a9ed | 1025 | } |
1da177e4 LT |
1026 | } |
1027 | ||
e3e32170 RJ |
1028 | success = icmp_pointers[icmph->type].handler(skb); |
1029 | ||
1030 | if (success) { | |
1031 | consume_skb(skb); | |
1032 | return 0; | |
1033 | } | |
1da177e4 LT |
1034 | |
1035 | drop: | |
1036 | kfree_skb(skb); | |
1037 | return 0; | |
6a5dc9e5 ED |
1038 | csum_error: |
1039 | ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS); | |
1da177e4 | 1040 | error: |
dcfc23ca | 1041 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
1042 | goto drop; |
1043 | } | |
1044 | ||
5b052042 LW |
1045 | void icmp_err(struct sk_buff *skb, u32 info) |
1046 | { | |
1047 | struct iphdr *iph = (struct iphdr *)skb->data; | |
6d0bfe22 LC |
1048 | int offset = iph->ihl<<2; |
1049 | struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset); | |
5b052042 LW |
1050 | int type = icmp_hdr(skb)->type; |
1051 | int code = icmp_hdr(skb)->code; | |
1052 | struct net *net = dev_net(skb->dev); | |
1053 | ||
1054 | /* | |
1055 | * Use ping_err to handle all icmp errors except those | |
1056 | * triggered by ICMP_ECHOREPLY which sent from kernel. | |
1057 | */ | |
1058 | if (icmph->type != ICMP_ECHOREPLY) { | |
6d0bfe22 | 1059 | ping_err(skb, offset, info); |
5b052042 LW |
1060 | return; |
1061 | } | |
1062 | ||
1063 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
1064 | ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0); | |
1065 | else if (type == ICMP_REDIRECT) | |
1066 | ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0); | |
1067 | } | |
1068 | ||
1da177e4 LT |
1069 | /* |
1070 | * This table is the definition of how we handle ICMP. | |
1071 | */ | |
9b5b5cff | 1072 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = { |
1da177e4 | 1073 | [ICMP_ECHOREPLY] = { |
c319b4d7 | 1074 | .handler = ping_rcv, |
1da177e4 LT |
1075 | }, |
1076 | [1] = { | |
1da177e4 LT |
1077 | .handler = icmp_discard, |
1078 | .error = 1, | |
1079 | }, | |
1080 | [2] = { | |
1da177e4 LT |
1081 | .handler = icmp_discard, |
1082 | .error = 1, | |
1083 | }, | |
1084 | [ICMP_DEST_UNREACH] = { | |
1da177e4 LT |
1085 | .handler = icmp_unreach, |
1086 | .error = 1, | |
1087 | }, | |
1088 | [ICMP_SOURCE_QUENCH] = { | |
1da177e4 LT |
1089 | .handler = icmp_unreach, |
1090 | .error = 1, | |
1091 | }, | |
1092 | [ICMP_REDIRECT] = { | |
1da177e4 LT |
1093 | .handler = icmp_redirect, |
1094 | .error = 1, | |
1095 | }, | |
1096 | [6] = { | |
1da177e4 LT |
1097 | .handler = icmp_discard, |
1098 | .error = 1, | |
1099 | }, | |
1100 | [7] = { | |
1da177e4 LT |
1101 | .handler = icmp_discard, |
1102 | .error = 1, | |
1103 | }, | |
1104 | [ICMP_ECHO] = { | |
1da177e4 LT |
1105 | .handler = icmp_echo, |
1106 | }, | |
1107 | [9] = { | |
1da177e4 LT |
1108 | .handler = icmp_discard, |
1109 | .error = 1, | |
1110 | }, | |
1111 | [10] = { | |
1da177e4 LT |
1112 | .handler = icmp_discard, |
1113 | .error = 1, | |
1114 | }, | |
1115 | [ICMP_TIME_EXCEEDED] = { | |
1da177e4 LT |
1116 | .handler = icmp_unreach, |
1117 | .error = 1, | |
1118 | }, | |
1119 | [ICMP_PARAMETERPROB] = { | |
1da177e4 LT |
1120 | .handler = icmp_unreach, |
1121 | .error = 1, | |
1122 | }, | |
1123 | [ICMP_TIMESTAMP] = { | |
1da177e4 LT |
1124 | .handler = icmp_timestamp, |
1125 | }, | |
1126 | [ICMP_TIMESTAMPREPLY] = { | |
1da177e4 LT |
1127 | .handler = icmp_discard, |
1128 | }, | |
1129 | [ICMP_INFO_REQUEST] = { | |
1da177e4 LT |
1130 | .handler = icmp_discard, |
1131 | }, | |
e905a9ed | 1132 | [ICMP_INFO_REPLY] = { |
1da177e4 LT |
1133 | .handler = icmp_discard, |
1134 | }, | |
1135 | [ICMP_ADDRESS] = { | |
838942a5 | 1136 | .handler = icmp_discard, |
1da177e4 LT |
1137 | }, |
1138 | [ICMP_ADDRESSREPLY] = { | |
838942a5 | 1139 | .handler = icmp_discard, |
1da177e4 LT |
1140 | }, |
1141 | }; | |
1142 | ||
4a6ad7a1 | 1143 | static void __net_exit icmp_sk_exit(struct net *net) |
1da177e4 | 1144 | { |
1da177e4 LT |
1145 | int i; |
1146 | ||
5c8cafd6 | 1147 | for_each_possible_cpu(i) |
349c9e3c ED |
1148 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i)); |
1149 | free_percpu(net->ipv4.icmp_sk); | |
4a6ad7a1 | 1150 | net->ipv4.icmp_sk = NULL; |
a5710d65 | 1151 | } |
1da177e4 | 1152 | |
263173af | 1153 | static int __net_init icmp_sk_init(struct net *net) |
a5710d65 | 1154 | { |
a5710d65 DL |
1155 | int i, err; |
1156 | ||
349c9e3c ED |
1157 | net->ipv4.icmp_sk = alloc_percpu(struct sock *); |
1158 | if (!net->ipv4.icmp_sk) | |
79c91159 DL |
1159 | return -ENOMEM; |
1160 | ||
a5710d65 | 1161 | for_each_possible_cpu(i) { |
1e3cf683 | 1162 | struct sock *sk; |
1da177e4 | 1163 | |
c1e9894d DL |
1164 | err = inet_ctl_sock_create(&sk, PF_INET, |
1165 | SOCK_RAW, IPPROTO_ICMP, net); | |
1da177e4 | 1166 | if (err < 0) |
a5710d65 | 1167 | goto fail; |
1da177e4 | 1168 | |
349c9e3c | 1169 | *per_cpu_ptr(net->ipv4.icmp_sk, i) = sk; |
1da177e4 LT |
1170 | |
1171 | /* Enough space for 2 64K ICMP packets, including | |
87fb4b7b | 1172 | * sk_buff/skb_shared_info struct overhead. |
1da177e4 | 1173 | */ |
87fb4b7b | 1174 | sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); |
1da177e4 | 1175 | |
b3a5b6cc ED |
1176 | /* |
1177 | * Speedup sock_wfree() | |
1178 | */ | |
1179 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
c1e9894d | 1180 | inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT; |
1da177e4 | 1181 | } |
a24022e1 PE |
1182 | |
1183 | /* Control parameters for ECHO replies. */ | |
1184 | net->ipv4.sysctl_icmp_echo_ignore_all = 0; | |
1185 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1; | |
1186 | ||
1187 | /* Control parameter - ignore bogus broadcast responses? */ | |
1188 | net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1; | |
1189 | ||
1190 | /* | |
1191 | * Configurable global rate limit. | |
1192 | * | |
1193 | * ratelimit defines tokens/packet consumed for dst->rate_token | |
1194 | * bucket ratemask defines which icmp types are ratelimited by | |
1195 | * setting it's bit position. | |
1196 | * | |
1197 | * default: | |
1198 | * dest unreachable (3), source quench (4), | |
1199 | * time exceeded (11), parameter problem (12) | |
1200 | */ | |
1201 | ||
1202 | net->ipv4.sysctl_icmp_ratelimit = 1 * HZ; | |
1203 | net->ipv4.sysctl_icmp_ratemask = 0x1818; | |
1204 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0; | |
1205 | ||
a5710d65 DL |
1206 | return 0; |
1207 | ||
1208 | fail: | |
1d1c8d13 | 1209 | for_each_possible_cpu(i) |
349c9e3c ED |
1210 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i)); |
1211 | free_percpu(net->ipv4.icmp_sk); | |
a5710d65 | 1212 | return err; |
1da177e4 LT |
1213 | } |
1214 | ||
4a6ad7a1 DL |
1215 | static struct pernet_operations __net_initdata icmp_sk_ops = { |
1216 | .init = icmp_sk_init, | |
1217 | .exit = icmp_sk_exit, | |
1218 | }; | |
1219 | ||
1220 | int __init icmp_init(void) | |
1221 | { | |
959d2726 | 1222 | return register_pernet_subsys(&icmp_sk_ops); |
4a6ad7a1 | 1223 | } |