]>
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> | |
7c0f6ba6 | 94 | #include <linux/uaccess.h> |
1da177e4 | 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> |
385add90 | 99 | #include <net/l3mdev.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 | |
7ba91ecb | 212 | /* Called with BH disabled */ |
fdc0bde9 | 213 | static inline struct sock *icmp_xmit_lock(struct net *net) |
1da177e4 | 214 | { |
fdc0bde9 DL |
215 | struct sock *sk; |
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 | */ | |
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 | { |
7ba91ecb | 230 | spin_unlock(&sk->sk_lock.slock); |
1da177e4 LT |
231 | } |
232 | ||
4cdf507d ED |
233 | int sysctl_icmp_msgs_per_sec __read_mostly = 1000; |
234 | int sysctl_icmp_msgs_burst __read_mostly = 50; | |
235 | ||
236 | static struct { | |
237 | spinlock_t lock; | |
238 | u32 credit; | |
239 | u32 stamp; | |
240 | } icmp_global = { | |
241 | .lock = __SPIN_LOCK_UNLOCKED(icmp_global.lock), | |
242 | }; | |
243 | ||
244 | /** | |
245 | * icmp_global_allow - Are we allowed to send one more ICMP message ? | |
246 | * | |
247 | * Uses a token bucket to limit our ICMP messages to sysctl_icmp_msgs_per_sec. | |
248 | * Returns false if we reached the limit and can not send another packet. | |
249 | * Note: called with BH disabled | |
250 | */ | |
251 | bool icmp_global_allow(void) | |
252 | { | |
253 | u32 credit, delta, incr = 0, now = (u32)jiffies; | |
254 | bool rc = false; | |
255 | ||
256 | /* Check if token bucket is empty and cannot be refilled | |
76f1c515 ED |
257 | * without taking the spinlock. The READ_ONCE() are paired |
258 | * with the following WRITE_ONCE() in this same function. | |
4cdf507d | 259 | */ |
76f1c515 ED |
260 | if (!READ_ONCE(icmp_global.credit)) { |
261 | delta = min_t(u32, now - READ_ONCE(icmp_global.stamp), HZ); | |
4cdf507d ED |
262 | if (delta < HZ / 50) |
263 | return false; | |
264 | } | |
265 | ||
266 | spin_lock(&icmp_global.lock); | |
267 | delta = min_t(u32, now - icmp_global.stamp, HZ); | |
268 | if (delta >= HZ / 50) { | |
269 | incr = sysctl_icmp_msgs_per_sec * delta / HZ ; | |
270 | if (incr) | |
76f1c515 | 271 | WRITE_ONCE(icmp_global.stamp, now); |
4cdf507d ED |
272 | } |
273 | credit = min_t(u32, icmp_global.credit + incr, sysctl_icmp_msgs_burst); | |
274 | if (credit) { | |
275 | credit--; | |
276 | rc = true; | |
277 | } | |
76f1c515 | 278 | WRITE_ONCE(icmp_global.credit, credit); |
4cdf507d ED |
279 | spin_unlock(&icmp_global.lock); |
280 | return rc; | |
281 | } | |
282 | EXPORT_SYMBOL(icmp_global_allow); | |
283 | ||
c0303efe JDB |
284 | static bool icmpv4_mask_allow(struct net *net, int type, int code) |
285 | { | |
286 | if (type > NR_ICMP_TYPES) | |
287 | return true; | |
288 | ||
289 | /* Don't limit PMTU discovery. */ | |
290 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
291 | return true; | |
292 | ||
293 | /* Limit if icmp type is enabled in ratemask. */ | |
294 | if (!((1 << type) & net->ipv4.sysctl_icmp_ratemask)) | |
295 | return true; | |
296 | ||
297 | return false; | |
298 | } | |
299 | ||
300 | static bool icmpv4_global_allow(struct net *net, int type, int code) | |
301 | { | |
302 | if (icmpv4_mask_allow(net, type, code)) | |
303 | return true; | |
304 | ||
305 | if (icmp_global_allow()) | |
306 | return true; | |
307 | ||
308 | return false; | |
309 | } | |
310 | ||
1da177e4 LT |
311 | /* |
312 | * Send an ICMP frame. | |
313 | */ | |
314 | ||
4cdf507d ED |
315 | static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt, |
316 | struct flowi4 *fl4, int type, int code) | |
1da177e4 | 317 | { |
d8d1f30b | 318 | struct dst_entry *dst = &rt->dst; |
c0303efe | 319 | struct inet_peer *peer; |
92d86829 | 320 | bool rc = true; |
c0303efe | 321 | int vif; |
1da177e4 | 322 | |
c0303efe | 323 | if (icmpv4_mask_allow(net, type, code)) |
1da177e4 LT |
324 | goto out; |
325 | ||
326 | /* No rate limit on loopback */ | |
327 | if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) | |
e905a9ed | 328 | goto out; |
1da177e4 | 329 | |
c0303efe JDB |
330 | vif = l3mdev_master_ifindex(dst->dev); |
331 | peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, vif, 1); | |
332 | rc = inet_peer_xrlim_allow(peer, net->ipv4.sysctl_icmp_ratelimit); | |
333 | if (peer) | |
334 | inet_putpeer(peer); | |
1da177e4 LT |
335 | out: |
336 | return rc; | |
337 | } | |
338 | ||
339 | /* | |
340 | * Maintain the counters used in the SNMP statistics for outgoing ICMP | |
341 | */ | |
0388b004 | 342 | void icmp_out_count(struct net *net, unsigned char type) |
1da177e4 | 343 | { |
903fc196 | 344 | ICMPMSGOUT_INC_STATS(net, type); |
75c939bb | 345 | ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS); |
1da177e4 LT |
346 | } |
347 | ||
348 | /* | |
349 | * Checksum each fragment, and on the first include the headers and final | |
350 | * checksum. | |
351 | */ | |
352 | static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd, | |
353 | struct sk_buff *skb) | |
354 | { | |
355 | struct icmp_bxm *icmp_param = (struct icmp_bxm *)from; | |
5f92a738 | 356 | __wsum csum; |
1da177e4 LT |
357 | |
358 | csum = skb_copy_and_csum_bits(icmp_param->skb, | |
359 | icmp_param->offset + offset, | |
360 | to, len, 0); | |
361 | ||
362 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
363 | if (icmp_pointers[icmp_param->data.icmph.type].error) | |
364 | nf_ct_attach(skb, icmp_param->skb); | |
365 | return 0; | |
366 | } | |
367 | ||
368 | static void icmp_push_reply(struct icmp_bxm *icmp_param, | |
77968b78 | 369 | struct flowi4 *fl4, |
2e77d89b | 370 | struct ipcm_cookie *ipc, struct rtable **rt) |
1da177e4 | 371 | { |
1e3cf683 | 372 | struct sock *sk; |
1da177e4 LT |
373 | struct sk_buff *skb; |
374 | ||
d8d1f30b | 375 | sk = icmp_sk(dev_net((*rt)->dst.dev)); |
f5fca608 | 376 | if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param, |
e905a9ed YH |
377 | icmp_param->data_len+icmp_param->head_len, |
378 | icmp_param->head_len, | |
1f8438a8 | 379 | ipc, rt, MSG_DONTWAIT) < 0) { |
5d3848bc | 380 | __ICMP_INC_STATS(sock_net(sk), ICMP_MIB_OUTERRORS); |
1e3cf683 | 381 | ip_flush_pending_frames(sk); |
1f8438a8 | 382 | } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) { |
88c7664f | 383 | struct icmphdr *icmph = icmp_hdr(skb); |
d3bc23e7 | 384 | __wsum csum = 0; |
1da177e4 LT |
385 | struct sk_buff *skb1; |
386 | ||
1e3cf683 | 387 | skb_queue_walk(&sk->sk_write_queue, skb1) { |
1da177e4 LT |
388 | csum = csum_add(csum, skb1->csum); |
389 | } | |
390 | csum = csum_partial_copy_nocheck((void *)&icmp_param->data, | |
391 | (char *)icmph, | |
392 | icmp_param->head_len, csum); | |
393 | icmph->checksum = csum_fold(csum); | |
394 | skb->ip_summed = CHECKSUM_NONE; | |
77968b78 | 395 | ip_push_pending_frames(sk, fl4); |
1da177e4 LT |
396 | } |
397 | } | |
398 | ||
399 | /* | |
400 | * Driving logic for building and sending ICMP messages. | |
401 | */ | |
402 | ||
403 | static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb) | |
404 | { | |
1da177e4 | 405 | struct ipcm_cookie ipc; |
511c3f92 | 406 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 407 | struct net *net = dev_net(rt->dst.dev); |
77968b78 | 408 | struct flowi4 fl4; |
fdc0bde9 DL |
409 | struct sock *sk; |
410 | struct inet_sock *inet; | |
35ebf65e | 411 | __be32 daddr, saddr; |
e110861f | 412 | u32 mark = IP4_REPLY_MARK(net, skb->mark); |
c0303efe JDB |
413 | int type = icmp_param->data.icmph.type; |
414 | int code = icmp_param->data.icmph.code; | |
1da177e4 | 415 | |
91ed1e66 | 416 | if (ip_options_echo(net, &icmp_param->replyopts.opt.opt, skb)) |
f00c401b | 417 | return; |
1da177e4 | 418 | |
7ba91ecb JDB |
419 | /* Needed by both icmp_global_allow and icmp_xmit_lock */ |
420 | local_bh_disable(); | |
1da177e4 | 421 | |
c0303efe JDB |
422 | /* global icmp_msgs_per_sec */ |
423 | if (!icmpv4_global_allow(net, type, code)) | |
7ba91ecb JDB |
424 | goto out_bh_enable; |
425 | ||
426 | sk = icmp_xmit_lock(net); | |
427 | if (!sk) | |
428 | goto out_bh_enable; | |
429 | inet = inet_sk(sk); | |
c0303efe | 430 | |
1da177e4 | 431 | icmp_param->data.icmph.checksum = 0; |
1da177e4 | 432 | |
eddc9ec5 | 433 | inet->tos = ip_hdr(skb)->tos; |
e110861f | 434 | sk->sk_mark = mark; |
9f6abb5f | 435 | daddr = ipc.addr = ip_hdr(skb)->saddr; |
35ebf65e | 436 | saddr = fib_compute_spec_dst(skb); |
1da177e4 | 437 | ipc.opt = NULL; |
2244d07b | 438 | ipc.tx_flags = 0; |
aa661581 FF |
439 | ipc.ttl = 0; |
440 | ipc.tos = -1; | |
441 | ||
f6d8bd05 ED |
442 | if (icmp_param->replyopts.opt.opt.optlen) { |
443 | ipc.opt = &icmp_param->replyopts.opt; | |
444 | if (ipc.opt->opt.srr) | |
445 | daddr = icmp_param->replyopts.opt.opt.faddr; | |
1da177e4 | 446 | } |
77968b78 DM |
447 | memset(&fl4, 0, sizeof(fl4)); |
448 | fl4.daddr = daddr; | |
35ebf65e | 449 | fl4.saddr = saddr; |
e110861f | 450 | fl4.flowi4_mark = mark; |
e2d118a1 | 451 | fl4.flowi4_uid = sock_net_uid(net, NULL); |
77968b78 DM |
452 | fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos); |
453 | fl4.flowi4_proto = IPPROTO_ICMP; | |
385add90 | 454 | fl4.flowi4_oif = l3mdev_master_ifindex(skb->dev); |
77968b78 DM |
455 | security_skb_classify_flow(skb, flowi4_to_flowi(&fl4)); |
456 | rt = ip_route_output_key(net, &fl4); | |
457 | if (IS_ERR(rt)) | |
458 | goto out_unlock; | |
c0303efe | 459 | if (icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
77968b78 | 460 | icmp_push_reply(icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
461 | ip_rt_put(rt); |
462 | out_unlock: | |
405666db | 463 | icmp_xmit_unlock(sk); |
7ba91ecb JDB |
464 | out_bh_enable: |
465 | local_bh_enable(); | |
1da177e4 LT |
466 | } |
467 | ||
77968b78 DM |
468 | static struct rtable *icmp_route_lookup(struct net *net, |
469 | struct flowi4 *fl4, | |
470 | struct sk_buff *skb_in, | |
b71d1d42 | 471 | const struct iphdr *iph, |
e110861f | 472 | __be32 saddr, u8 tos, u32 mark, |
f6d460cf DM |
473 | int type, int code, |
474 | struct icmp_bxm *param) | |
475 | { | |
f6d460cf | 476 | struct rtable *rt, *rt2; |
415b3334 | 477 | struct flowi4 fl4_dec; |
f6d460cf DM |
478 | int err; |
479 | ||
77968b78 DM |
480 | memset(fl4, 0, sizeof(*fl4)); |
481 | fl4->daddr = (param->replyopts.opt.opt.srr ? | |
482 | param->replyopts.opt.opt.faddr : iph->saddr); | |
483 | fl4->saddr = saddr; | |
e110861f | 484 | fl4->flowi4_mark = mark; |
e2d118a1 | 485 | fl4->flowi4_uid = sock_net_uid(net, NULL); |
77968b78 DM |
486 | fl4->flowi4_tos = RT_TOS(tos); |
487 | fl4->flowi4_proto = IPPROTO_ICMP; | |
488 | fl4->fl4_icmp_type = type; | |
489 | fl4->fl4_icmp_code = code; | |
9d1a6c4e | 490 | fl4->flowi4_oif = l3mdev_master_ifindex(skb_dst(skb_in)->dev); |
613d09b3 | 491 | |
77968b78 | 492 | security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4)); |
3abd1ade | 493 | rt = ip_route_output_key_hash(net, fl4, skb_in); |
b23dd4fe DM |
494 | if (IS_ERR(rt)) |
495 | return rt; | |
f6d460cf DM |
496 | |
497 | /* No need to clone since we're just using its address. */ | |
498 | rt2 = rt; | |
499 | ||
9d6ec938 | 500 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
77968b78 | 501 | flowi4_to_flowi(fl4), NULL, 0); |
452edd59 | 502 | if (!IS_ERR(rt)) { |
f6d460cf DM |
503 | if (rt != rt2) |
504 | return rt; | |
452edd59 | 505 | } else if (PTR_ERR(rt) == -EPERM) { |
f6d460cf | 506 | rt = NULL; |
452edd59 DM |
507 | } else |
508 | return rt; | |
f6d460cf | 509 | |
415b3334 | 510 | err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET); |
f6d460cf DM |
511 | if (err) |
512 | goto relookup_failed; | |
513 | ||
9d1a6c4e | 514 | if (inet_addr_type_dev_table(net, skb_dst(skb_in)->dev, |
30bbaa19 | 515 | fl4_dec.saddr) == RTN_LOCAL) { |
415b3334 | 516 | rt2 = __ip_route_output_key(net, &fl4_dec); |
b23dd4fe DM |
517 | if (IS_ERR(rt2)) |
518 | err = PTR_ERR(rt2); | |
f6d460cf | 519 | } else { |
9d6ec938 | 520 | struct flowi4 fl4_2 = {}; |
f6d460cf DM |
521 | unsigned long orefdst; |
522 | ||
415b3334 | 523 | fl4_2.daddr = fl4_dec.saddr; |
9d6ec938 | 524 | rt2 = ip_route_output_key(net, &fl4_2); |
b23dd4fe DM |
525 | if (IS_ERR(rt2)) { |
526 | err = PTR_ERR(rt2); | |
f6d460cf | 527 | goto relookup_failed; |
b23dd4fe | 528 | } |
f6d460cf DM |
529 | /* Ugh! */ |
530 | orefdst = skb_in->_skb_refdst; /* save old refdst */ | |
773a69d6 | 531 | skb_dst_set(skb_in, NULL); |
415b3334 | 532 | err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr, |
f6d460cf DM |
533 | RT_TOS(tos), rt2->dst.dev); |
534 | ||
535 | dst_release(&rt2->dst); | |
536 | rt2 = skb_rtable(skb_in); | |
537 | skb_in->_skb_refdst = orefdst; /* restore old refdst */ | |
538 | } | |
539 | ||
540 | if (err) | |
541 | goto relookup_failed; | |
542 | ||
9d6ec938 | 543 | rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst, |
415b3334 | 544 | flowi4_to_flowi(&fl4_dec), NULL, |
9d6ec938 | 545 | XFRM_LOOKUP_ICMP); |
452edd59 | 546 | if (!IS_ERR(rt2)) { |
f6d460cf | 547 | dst_release(&rt->dst); |
415b3334 | 548 | memcpy(fl4, &fl4_dec, sizeof(*fl4)); |
f6d460cf | 549 | rt = rt2; |
452edd59 DM |
550 | } else if (PTR_ERR(rt2) == -EPERM) { |
551 | if (rt) | |
552 | dst_release(&rt->dst); | |
553 | return rt2; | |
554 | } else { | |
555 | err = PTR_ERR(rt2); | |
556 | goto relookup_failed; | |
f6d460cf | 557 | } |
f6d460cf DM |
558 | return rt; |
559 | ||
560 | relookup_failed: | |
561 | if (rt) | |
562 | return rt; | |
563 | return ERR_PTR(err); | |
564 | } | |
1da177e4 LT |
565 | |
566 | /* | |
567 | * Send an ICMP message in response to a situation | |
568 | * | |
569 | * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header. | |
570 | * MAY send more (we do). | |
571 | * MUST NOT change this header information. | |
572 | * MUST NOT reply to a multicast/broadcast IP address. | |
573 | * MUST NOT reply to a multicast/broadcast MAC address. | |
574 | * MUST reply to only the first fragment. | |
575 | */ | |
576 | ||
17fa6c95 NS |
577 | void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info, |
578 | const struct ip_options *opt) | |
1da177e4 LT |
579 | { |
580 | struct iphdr *iph; | |
581 | int room; | |
8d9ba388 | 582 | struct icmp_bxm icmp_param; |
511c3f92 | 583 | struct rtable *rt = skb_rtable(skb_in); |
1da177e4 | 584 | struct ipcm_cookie ipc; |
77968b78 | 585 | struct flowi4 fl4; |
a61ced5d | 586 | __be32 saddr; |
1da177e4 | 587 | u8 tos; |
e110861f | 588 | u32 mark; |
dde1bc0e | 589 | struct net *net; |
4a6ad7a1 | 590 | struct sock *sk; |
1da177e4 LT |
591 | |
592 | if (!rt) | |
593 | goto out; | |
8bfbf04a HL |
594 | |
595 | if (rt->dst.dev) | |
596 | net = dev_net(rt->dst.dev); | |
597 | else if (skb_in->dev) | |
598 | net = dev_net(skb_in->dev); | |
599 | else | |
600 | goto out; | |
1da177e4 LT |
601 | |
602 | /* | |
603 | * Find the original header. It is expected to be valid, of course. | |
604 | * Check this, icmp_send is called from the most obscure devices | |
605 | * sometimes. | |
606 | */ | |
eddc9ec5 | 607 | iph = ip_hdr(skb_in); |
1da177e4 | 608 | |
27a884dc | 609 | if ((u8 *)iph < skb_in->head || |
f7c0c2ae SH |
610 | (skb_network_header(skb_in) + sizeof(*iph)) > |
611 | skb_tail_pointer(skb_in)) | |
1da177e4 LT |
612 | goto out; |
613 | ||
614 | /* | |
615 | * No replies to physical multicast/broadcast | |
616 | */ | |
617 | if (skb_in->pkt_type != PACKET_HOST) | |
618 | goto out; | |
619 | ||
620 | /* | |
621 | * Now check at the protocol level | |
622 | */ | |
623 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
624 | goto out; | |
625 | ||
626 | /* | |
627 | * Only reply to fragment 0. We byte re-order the constant | |
628 | * mask for efficiency. | |
629 | */ | |
630 | if (iph->frag_off & htons(IP_OFFSET)) | |
631 | goto out; | |
632 | ||
633 | /* | |
634 | * If we send an ICMP error to an ICMP error a mess would result.. | |
635 | */ | |
636 | if (icmp_pointers[type].error) { | |
637 | /* | |
638 | * We are an error, check if we are replying to an | |
639 | * ICMP error | |
640 | */ | |
641 | if (iph->protocol == IPPROTO_ICMP) { | |
642 | u8 _inner_type, *itp; | |
643 | ||
644 | itp = skb_header_pointer(skb_in, | |
d56f90a7 | 645 | skb_network_header(skb_in) + |
1da177e4 LT |
646 | (iph->ihl << 2) + |
647 | offsetof(struct icmphdr, | |
648 | type) - | |
649 | skb_in->data, | |
650 | sizeof(_inner_type), | |
651 | &_inner_type); | |
51456b29 | 652 | if (!itp) |
1da177e4 LT |
653 | goto out; |
654 | ||
655 | /* | |
656 | * Assume any unknown ICMP type is an error. This | |
657 | * isn't specified by the RFC, but think about it.. | |
658 | */ | |
659 | if (*itp > NR_ICMP_TYPES || | |
660 | icmp_pointers[*itp].error) | |
661 | goto out; | |
662 | } | |
663 | } | |
664 | ||
7ba91ecb JDB |
665 | /* Needed by both icmp_global_allow and icmp_xmit_lock */ |
666 | local_bh_disable(); | |
c0303efe | 667 | |
849a44de JDB |
668 | /* Check global sysctl_icmp_msgs_per_sec ratelimit, unless |
669 | * incoming dev is loopback. If outgoing dev change to not be | |
670 | * loopback, then peer ratelimit still work (in icmpv4_xrlim_allow) | |
671 | */ | |
672 | if (!(skb_in->dev && (skb_in->dev->flags&IFF_LOOPBACK)) && | |
673 | !icmpv4_global_allow(net, type, code)) | |
7ba91ecb JDB |
674 | goto out_bh_enable; |
675 | ||
676 | sk = icmp_xmit_lock(net); | |
677 | if (!sk) | |
678 | goto out_bh_enable; | |
1da177e4 LT |
679 | |
680 | /* | |
681 | * Construct source address and options. | |
682 | */ | |
683 | ||
684 | saddr = iph->daddr; | |
02a6d613 | 685 | if (!(rt->rt_flags & RTCF_LOCAL)) { |
6e1d9103 PM |
686 | struct net_device *dev = NULL; |
687 | ||
685c7944 | 688 | rcu_read_lock(); |
c7537967 DM |
689 | if (rt_is_input_route(rt) && |
690 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr) | |
92101b3b | 691 | dev = dev_get_by_index_rcu(net, inet_iif(skb_in)); |
6e1d9103 | 692 | |
685c7944 | 693 | if (dev) |
6e1d9103 | 694 | saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
685c7944 | 695 | else |
1c2fb7f9 | 696 | saddr = 0; |
685c7944 | 697 | rcu_read_unlock(); |
1c2fb7f9 | 698 | } |
1da177e4 LT |
699 | |
700 | tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) | | |
701 | IPTOS_PREC_INTERNETCONTROL) : | |
702 | iph->tos; | |
e110861f | 703 | mark = IP4_REPLY_MARK(net, skb_in->mark); |
1da177e4 | 704 | |
17fa6c95 | 705 | if (__ip_options_echo(net, &icmp_param.replyopts.opt.opt, skb_in, opt)) |
fa60cf7f | 706 | goto out_unlock; |
1da177e4 LT |
707 | |
708 | ||
709 | /* | |
710 | * Prepare data for ICMP header. | |
711 | */ | |
712 | ||
8d9ba388 JDB |
713 | icmp_param.data.icmph.type = type; |
714 | icmp_param.data.icmph.code = code; | |
715 | icmp_param.data.icmph.un.gateway = info; | |
716 | icmp_param.data.icmph.checksum = 0; | |
717 | icmp_param.skb = skb_in; | |
718 | icmp_param.offset = skb_network_offset(skb_in); | |
405666db | 719 | inet_sk(sk)->tos = tos; |
e110861f | 720 | sk->sk_mark = mark; |
1da177e4 | 721 | ipc.addr = iph->saddr; |
8d9ba388 | 722 | ipc.opt = &icmp_param.replyopts.opt; |
2244d07b | 723 | ipc.tx_flags = 0; |
aa661581 FF |
724 | ipc.ttl = 0; |
725 | ipc.tos = -1; | |
1da177e4 | 726 | |
e110861f | 727 | rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, mark, |
8d9ba388 | 728 | type, code, &icmp_param); |
f6d460cf DM |
729 | if (IS_ERR(rt)) |
730 | goto out_unlock; | |
1da177e4 | 731 | |
c0303efe | 732 | /* peer icmp_ratelimit */ |
a48eff12 | 733 | if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
1da177e4 LT |
734 | goto ende; |
735 | ||
736 | /* RFC says return as much as we can without exceeding 576 bytes. */ | |
737 | ||
d8d1f30b | 738 | room = dst_mtu(&rt->dst); |
1da177e4 LT |
739 | if (room > 576) |
740 | room = 576; | |
8d9ba388 | 741 | room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen; |
1da177e4 LT |
742 | room -= sizeof(struct icmphdr); |
743 | ||
8d9ba388 JDB |
744 | icmp_param.data_len = skb_in->len - icmp_param.offset; |
745 | if (icmp_param.data_len > room) | |
746 | icmp_param.data_len = room; | |
747 | icmp_param.head_len = sizeof(struct icmphdr); | |
1da177e4 | 748 | |
8d9ba388 | 749 | icmp_push_reply(&icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
750 | ende: |
751 | ip_rt_put(rt); | |
752 | out_unlock: | |
405666db | 753 | icmp_xmit_unlock(sk); |
7ba91ecb JDB |
754 | out_bh_enable: |
755 | local_bh_enable(); | |
1da177e4 LT |
756 | out:; |
757 | } | |
17fa6c95 | 758 | EXPORT_SYMBOL(__icmp_send); |
1da177e4 LT |
759 | |
760 | ||
1de9243b DM |
761 | static void icmp_socket_deliver(struct sk_buff *skb, u32 info) |
762 | { | |
763 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
764 | const struct net_protocol *ipprot; | |
765 | int protocol = iph->protocol; | |
766 | ||
f0a70e90 DM |
767 | /* Checkin full IP header plus 8 bytes of protocol to |
768 | * avoid additional coding at protocol handlers. | |
769 | */ | |
7304fe46 | 770 | if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) { |
5d3848bc | 771 | __ICMP_INC_STATS(dev_net(skb->dev), ICMP_MIB_INERRORS); |
f0a70e90 | 772 | return; |
7304fe46 | 773 | } |
f0a70e90 | 774 | |
1de9243b DM |
775 | raw_icmp_error(skb, protocol, info); |
776 | ||
1de9243b DM |
777 | ipprot = rcu_dereference(inet_protos[protocol]); |
778 | if (ipprot && ipprot->err_handler) | |
779 | ipprot->err_handler(skb, info); | |
1de9243b DM |
780 | } |
781 | ||
8ed1dc44 HFS |
782 | static bool icmp_tag_validation(int proto) |
783 | { | |
784 | bool ok; | |
785 | ||
786 | rcu_read_lock(); | |
787 | ok = rcu_dereference(inet_protos[proto])->icmp_strict_tag_validation; | |
788 | rcu_read_unlock(); | |
789 | return ok; | |
790 | } | |
791 | ||
1da177e4 | 792 | /* |
258bbb1b | 793 | * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEEDED, ICMP_QUENCH, and |
08578d8d | 794 | * ICMP_PARAMETERPROB. |
1da177e4 LT |
795 | */ |
796 | ||
e3e32170 | 797 | static bool icmp_unreach(struct sk_buff *skb) |
1da177e4 | 798 | { |
b71d1d42 | 799 | const struct iphdr *iph; |
1da177e4 | 800 | struct icmphdr *icmph; |
dde1bc0e | 801 | struct net *net; |
f9242b6b | 802 | u32 info = 0; |
dde1bc0e | 803 | |
adf30907 | 804 | net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
805 | |
806 | /* | |
807 | * Incomplete header ? | |
808 | * Only checks for the IP header, there should be an | |
809 | * additional check for longer headers in upper levels. | |
810 | */ | |
811 | ||
812 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
813 | goto out_err; | |
814 | ||
88c7664f | 815 | icmph = icmp_hdr(skb); |
b71d1d42 | 816 | iph = (const struct iphdr *)skb->data; |
1da177e4 LT |
817 | |
818 | if (iph->ihl < 5) /* Mangled header, drop. */ | |
819 | goto out_err; | |
820 | ||
258bbb1b MC |
821 | switch (icmph->type) { |
822 | case ICMP_DEST_UNREACH: | |
1da177e4 LT |
823 | switch (icmph->code & 15) { |
824 | case ICMP_NET_UNREACH: | |
825 | case ICMP_HOST_UNREACH: | |
826 | case ICMP_PROT_UNREACH: | |
827 | case ICMP_PORT_UNREACH: | |
828 | break; | |
829 | case ICMP_FRAG_NEEDED: | |
8ed1dc44 HFS |
830 | /* for documentation of the ip_no_pmtu_disc |
831 | * values please see | |
832 | * Documentation/networking/ip-sysctl.txt | |
833 | */ | |
834 | switch (net->ipv4.sysctl_ip_no_pmtu_disc) { | |
835 | default: | |
ba7a46f1 JP |
836 | net_dbg_ratelimited("%pI4: fragmentation needed and DF set\n", |
837 | &iph->daddr); | |
8ed1dc44 HFS |
838 | break; |
839 | case 2: | |
840 | goto out; | |
841 | case 3: | |
842 | if (!icmp_tag_validation(iph->protocol)) | |
843 | goto out; | |
844 | /* fall through */ | |
845 | case 0: | |
46517008 | 846 | info = ntohs(icmph->un.frag.mtu); |
1da177e4 LT |
847 | } |
848 | break; | |
849 | case ICMP_SR_FAILED: | |
ba7a46f1 JP |
850 | net_dbg_ratelimited("%pI4: Source Route Failed\n", |
851 | &iph->daddr); | |
1da177e4 LT |
852 | break; |
853 | default: | |
854 | break; | |
855 | } | |
856 | if (icmph->code > NR_ICMP_UNREACH) | |
857 | goto out; | |
258bbb1b MC |
858 | break; |
859 | case ICMP_PARAMETERPROB: | |
1da177e4 | 860 | info = ntohl(icmph->un.gateway) >> 24; |
258bbb1b MC |
861 | break; |
862 | case ICMP_TIME_EXCEEDED: | |
863 | __ICMP_INC_STATS(net, ICMP_MIB_INTIMEEXCDS); | |
864 | if (icmph->code == ICMP_EXC_FRAGTIME) | |
865 | goto out; | |
866 | break; | |
867 | } | |
1da177e4 LT |
868 | |
869 | /* | |
870 | * Throw it at our lower layers | |
871 | * | |
872 | * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed | |
873 | * header. | |
874 | * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the | |
875 | * transport layer. | |
876 | * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to | |
877 | * transport layer. | |
878 | */ | |
879 | ||
880 | /* | |
25985edc | 881 | * Check the other end isn't violating RFC 1122. Some routers send |
1da177e4 LT |
882 | * bogus responses to broadcast frames. If you see this message |
883 | * first check your netmask matches at both ends, if it does then | |
884 | * get the other vendor to fix their kit. | |
885 | */ | |
886 | ||
b34a95ee | 887 | if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses && |
30bbaa19 | 888 | inet_addr_type_dev_table(net, skb->dev, iph->daddr) == RTN_BROADCAST) { |
e87cc472 JP |
889 | net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n", |
890 | &ip_hdr(skb)->saddr, | |
891 | icmph->type, icmph->code, | |
892 | &iph->daddr, skb->dev->name); | |
1da177e4 LT |
893 | goto out; |
894 | } | |
895 | ||
1de9243b | 896 | icmp_socket_deliver(skb, info); |
1da177e4 LT |
897 | |
898 | out: | |
e3e32170 | 899 | return true; |
1da177e4 | 900 | out_err: |
5d3848bc | 901 | __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); |
e3e32170 | 902 | return false; |
1da177e4 LT |
903 | } |
904 | ||
905 | ||
906 | /* | |
907 | * Handle ICMP_REDIRECT. | |
908 | */ | |
909 | ||
e3e32170 | 910 | static bool icmp_redirect(struct sk_buff *skb) |
1da177e4 | 911 | { |
94206125 | 912 | if (skb->len < sizeof(struct iphdr)) { |
5d3848bc | 913 | __ICMP_INC_STATS(dev_net(skb->dev), ICMP_MIB_INERRORS); |
e3e32170 | 914 | return false; |
94206125 | 915 | } |
1da177e4 | 916 | |
e3e32170 RJ |
917 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) { |
918 | /* there aught to be a stat */ | |
919 | return false; | |
920 | } | |
c319b4d7 | 921 | |
d3351b75 | 922 | icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway); |
e3e32170 | 923 | return true; |
1da177e4 LT |
924 | } |
925 | ||
926 | /* | |
927 | * Handle ICMP_ECHO ("ping") requests. | |
928 | * | |
929 | * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo | |
930 | * requests. | |
931 | * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be | |
932 | * included in the reply. | |
933 | * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring | |
934 | * echo requests, MUST have default=NOT. | |
935 | * See also WRT handling of options once they are done and working. | |
936 | */ | |
937 | ||
e3e32170 | 938 | static bool icmp_echo(struct sk_buff *skb) |
1da177e4 | 939 | { |
b34a95ee PE |
940 | struct net *net; |
941 | ||
adf30907 | 942 | net = dev_net(skb_dst(skb)->dev); |
b34a95ee | 943 | if (!net->ipv4.sysctl_icmp_echo_ignore_all) { |
1da177e4 LT |
944 | struct icmp_bxm icmp_param; |
945 | ||
88c7664f | 946 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
947 | icmp_param.data.icmph.type = ICMP_ECHOREPLY; |
948 | icmp_param.skb = skb; | |
949 | icmp_param.offset = 0; | |
950 | icmp_param.data_len = skb->len; | |
951 | icmp_param.head_len = sizeof(struct icmphdr); | |
952 | icmp_reply(&icmp_param, skb); | |
953 | } | |
e3e32170 RJ |
954 | /* should there be an ICMP stat for ignored echos? */ |
955 | return true; | |
1da177e4 LT |
956 | } |
957 | ||
958 | /* | |
959 | * Handle ICMP Timestamp requests. | |
960 | * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests. | |
961 | * SHOULD be in the kernel for minimum random latency. | |
962 | * MUST be accurate to a few minutes. | |
963 | * MUST be updated at least at 15Hz. | |
964 | */ | |
e3e32170 | 965 | static bool icmp_timestamp(struct sk_buff *skb) |
1da177e4 | 966 | { |
1da177e4 LT |
967 | struct icmp_bxm icmp_param; |
968 | /* | |
969 | * Too short. | |
970 | */ | |
971 | if (skb->len < 4) | |
972 | goto out_err; | |
973 | ||
974 | /* | |
975 | * Fill in the current time as ms since midnight UT: | |
976 | */ | |
822c8685 | 977 | icmp_param.data.times[1] = inet_current_timestamp(); |
1da177e4 | 978 | icmp_param.data.times[2] = icmp_param.data.times[1]; |
15285402 GS |
979 | |
980 | BUG_ON(skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4)); | |
981 | ||
88c7664f | 982 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
983 | icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY; |
984 | icmp_param.data.icmph.code = 0; | |
985 | icmp_param.skb = skb; | |
986 | icmp_param.offset = 0; | |
987 | icmp_param.data_len = 0; | |
988 | icmp_param.head_len = sizeof(struct icmphdr) + 12; | |
989 | icmp_reply(&icmp_param, skb); | |
e3e32170 RJ |
990 | return true; |
991 | ||
1da177e4 | 992 | out_err: |
5d3848bc | 993 | __ICMP_INC_STATS(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS); |
e3e32170 | 994 | return false; |
1da177e4 LT |
995 | } |
996 | ||
e3e32170 | 997 | static bool icmp_discard(struct sk_buff *skb) |
1da177e4 | 998 | { |
e3e32170 RJ |
999 | /* pretend it was a success */ |
1000 | return true; | |
1da177e4 LT |
1001 | } |
1002 | ||
1003 | /* | |
1004 | * Deal with incoming ICMP packets. | |
1005 | */ | |
1006 | int icmp_rcv(struct sk_buff *skb) | |
1007 | { | |
1008 | struct icmphdr *icmph; | |
511c3f92 | 1009 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 1010 | struct net *net = dev_net(rt->dst.dev); |
e3e32170 | 1011 | bool success; |
1da177e4 | 1012 | |
aebcf82c | 1013 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { |
def8b4fa | 1014 | struct sec_path *sp = skb_sec_path(skb); |
8b7817f3 HX |
1015 | int nh; |
1016 | ||
def8b4fa | 1017 | if (!(sp && sp->xvec[sp->len - 1]->props.flags & |
aebcf82c HX |
1018 | XFRM_STATE_ICMP)) |
1019 | goto drop; | |
1020 | ||
8b7817f3 HX |
1021 | if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr))) |
1022 | goto drop; | |
1023 | ||
1024 | nh = skb_network_offset(skb); | |
1025 | skb_set_network_header(skb, sizeof(*icmph)); | |
1026 | ||
1027 | if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) | |
1028 | goto drop; | |
1029 | ||
1030 | skb_set_network_header(skb, nh); | |
1031 | } | |
1032 | ||
5d3848bc | 1033 | __ICMP_INC_STATS(net, ICMP_MIB_INMSGS); |
1da177e4 | 1034 | |
29a96e1f TH |
1035 | if (skb_checksum_simple_validate(skb)) |
1036 | goto csum_error; | |
1da177e4 | 1037 | |
8cf22943 HX |
1038 | if (!pskb_pull(skb, sizeof(*icmph))) |
1039 | goto error; | |
1da177e4 | 1040 | |
88c7664f | 1041 | icmph = icmp_hdr(skb); |
1da177e4 | 1042 | |
214d3f1f | 1043 | ICMPMSGIN_INC_STATS(net, icmph->type); |
1da177e4 LT |
1044 | /* |
1045 | * 18 is the highest 'known' ICMP type. Anything else is a mystery | |
1046 | * | |
1047 | * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently | |
1048 | * discarded. | |
1049 | */ | |
1050 | if (icmph->type > NR_ICMP_TYPES) | |
1051 | goto error; | |
1052 | ||
1053 | ||
1054 | /* | |
1055 | * Parse the ICMP message | |
1056 | */ | |
1057 | ||
e905a9ed | 1058 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
1da177e4 LT |
1059 | /* |
1060 | * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be | |
1061 | * silently ignored (we let user decide with a sysctl). | |
1062 | * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently | |
1063 | * discarded if to broadcast/multicast. | |
1064 | */ | |
4c866aa7 AK |
1065 | if ((icmph->type == ICMP_ECHO || |
1066 | icmph->type == ICMP_TIMESTAMP) && | |
b34a95ee | 1067 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts) { |
1da177e4 LT |
1068 | goto error; |
1069 | } | |
1070 | if (icmph->type != ICMP_ECHO && | |
1071 | icmph->type != ICMP_TIMESTAMP && | |
1072 | icmph->type != ICMP_ADDRESS && | |
1073 | icmph->type != ICMP_ADDRESSREPLY) { | |
1074 | goto error; | |
e905a9ed | 1075 | } |
1da177e4 LT |
1076 | } |
1077 | ||
e3e32170 RJ |
1078 | success = icmp_pointers[icmph->type].handler(skb); |
1079 | ||
1080 | if (success) { | |
1081 | consume_skb(skb); | |
f91c58d6 | 1082 | return NET_RX_SUCCESS; |
e3e32170 | 1083 | } |
1da177e4 LT |
1084 | |
1085 | drop: | |
1086 | kfree_skb(skb); | |
f91c58d6 | 1087 | return NET_RX_DROP; |
6a5dc9e5 | 1088 | csum_error: |
5d3848bc | 1089 | __ICMP_INC_STATS(net, ICMP_MIB_CSUMERRORS); |
1da177e4 | 1090 | error: |
5d3848bc | 1091 | __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
1092 | goto drop; |
1093 | } | |
1094 | ||
5b052042 LW |
1095 | void icmp_err(struct sk_buff *skb, u32 info) |
1096 | { | |
1097 | struct iphdr *iph = (struct iphdr *)skb->data; | |
6d0bfe22 LC |
1098 | int offset = iph->ihl<<2; |
1099 | struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset); | |
5b052042 LW |
1100 | int type = icmp_hdr(skb)->type; |
1101 | int code = icmp_hdr(skb)->code; | |
1102 | struct net *net = dev_net(skb->dev); | |
1103 | ||
1104 | /* | |
1105 | * Use ping_err to handle all icmp errors except those | |
1106 | * triggered by ICMP_ECHOREPLY which sent from kernel. | |
1107 | */ | |
1108 | if (icmph->type != ICMP_ECHOREPLY) { | |
6d0bfe22 | 1109 | ping_err(skb, offset, info); |
5b052042 LW |
1110 | return; |
1111 | } | |
1112 | ||
1113 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
1114 | ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0); | |
1115 | else if (type == ICMP_REDIRECT) | |
1116 | ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0); | |
1117 | } | |
1118 | ||
1da177e4 LT |
1119 | /* |
1120 | * This table is the definition of how we handle ICMP. | |
1121 | */ | |
9b5b5cff | 1122 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = { |
1da177e4 | 1123 | [ICMP_ECHOREPLY] = { |
c319b4d7 | 1124 | .handler = ping_rcv, |
1da177e4 LT |
1125 | }, |
1126 | [1] = { | |
1da177e4 LT |
1127 | .handler = icmp_discard, |
1128 | .error = 1, | |
1129 | }, | |
1130 | [2] = { | |
1da177e4 LT |
1131 | .handler = icmp_discard, |
1132 | .error = 1, | |
1133 | }, | |
1134 | [ICMP_DEST_UNREACH] = { | |
1da177e4 LT |
1135 | .handler = icmp_unreach, |
1136 | .error = 1, | |
1137 | }, | |
1138 | [ICMP_SOURCE_QUENCH] = { | |
1da177e4 LT |
1139 | .handler = icmp_unreach, |
1140 | .error = 1, | |
1141 | }, | |
1142 | [ICMP_REDIRECT] = { | |
1da177e4 LT |
1143 | .handler = icmp_redirect, |
1144 | .error = 1, | |
1145 | }, | |
1146 | [6] = { | |
1da177e4 LT |
1147 | .handler = icmp_discard, |
1148 | .error = 1, | |
1149 | }, | |
1150 | [7] = { | |
1da177e4 LT |
1151 | .handler = icmp_discard, |
1152 | .error = 1, | |
1153 | }, | |
1154 | [ICMP_ECHO] = { | |
1da177e4 LT |
1155 | .handler = icmp_echo, |
1156 | }, | |
1157 | [9] = { | |
1da177e4 LT |
1158 | .handler = icmp_discard, |
1159 | .error = 1, | |
1160 | }, | |
1161 | [10] = { | |
1da177e4 LT |
1162 | .handler = icmp_discard, |
1163 | .error = 1, | |
1164 | }, | |
1165 | [ICMP_TIME_EXCEEDED] = { | |
1da177e4 LT |
1166 | .handler = icmp_unreach, |
1167 | .error = 1, | |
1168 | }, | |
1169 | [ICMP_PARAMETERPROB] = { | |
1da177e4 LT |
1170 | .handler = icmp_unreach, |
1171 | .error = 1, | |
1172 | }, | |
1173 | [ICMP_TIMESTAMP] = { | |
1da177e4 LT |
1174 | .handler = icmp_timestamp, |
1175 | }, | |
1176 | [ICMP_TIMESTAMPREPLY] = { | |
1da177e4 LT |
1177 | .handler = icmp_discard, |
1178 | }, | |
1179 | [ICMP_INFO_REQUEST] = { | |
1da177e4 LT |
1180 | .handler = icmp_discard, |
1181 | }, | |
e905a9ed | 1182 | [ICMP_INFO_REPLY] = { |
1da177e4 LT |
1183 | .handler = icmp_discard, |
1184 | }, | |
1185 | [ICMP_ADDRESS] = { | |
838942a5 | 1186 | .handler = icmp_discard, |
1da177e4 LT |
1187 | }, |
1188 | [ICMP_ADDRESSREPLY] = { | |
838942a5 | 1189 | .handler = icmp_discard, |
1da177e4 LT |
1190 | }, |
1191 | }; | |
1192 | ||
4a6ad7a1 | 1193 | static void __net_exit icmp_sk_exit(struct net *net) |
1da177e4 | 1194 | { |
1da177e4 LT |
1195 | int i; |
1196 | ||
5c8cafd6 | 1197 | for_each_possible_cpu(i) |
349c9e3c ED |
1198 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i)); |
1199 | free_percpu(net->ipv4.icmp_sk); | |
4a6ad7a1 | 1200 | net->ipv4.icmp_sk = NULL; |
a5710d65 | 1201 | } |
1da177e4 | 1202 | |
263173af | 1203 | static int __net_init icmp_sk_init(struct net *net) |
a5710d65 | 1204 | { |
a5710d65 DL |
1205 | int i, err; |
1206 | ||
349c9e3c ED |
1207 | net->ipv4.icmp_sk = alloc_percpu(struct sock *); |
1208 | if (!net->ipv4.icmp_sk) | |
79c91159 DL |
1209 | return -ENOMEM; |
1210 | ||
a5710d65 | 1211 | for_each_possible_cpu(i) { |
1e3cf683 | 1212 | struct sock *sk; |
1da177e4 | 1213 | |
c1e9894d DL |
1214 | err = inet_ctl_sock_create(&sk, PF_INET, |
1215 | SOCK_RAW, IPPROTO_ICMP, net); | |
1da177e4 | 1216 | if (err < 0) |
a5710d65 | 1217 | goto fail; |
1da177e4 | 1218 | |
349c9e3c | 1219 | *per_cpu_ptr(net->ipv4.icmp_sk, i) = sk; |
1da177e4 LT |
1220 | |
1221 | /* Enough space for 2 64K ICMP packets, including | |
87fb4b7b | 1222 | * sk_buff/skb_shared_info struct overhead. |
1da177e4 | 1223 | */ |
87fb4b7b | 1224 | sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); |
1da177e4 | 1225 | |
b3a5b6cc ED |
1226 | /* |
1227 | * Speedup sock_wfree() | |
1228 | */ | |
1229 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
c1e9894d | 1230 | inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT; |
1da177e4 | 1231 | } |
a24022e1 PE |
1232 | |
1233 | /* Control parameters for ECHO replies. */ | |
1234 | net->ipv4.sysctl_icmp_echo_ignore_all = 0; | |
1235 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1; | |
1236 | ||
1237 | /* Control parameter - ignore bogus broadcast responses? */ | |
1238 | net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1; | |
1239 | ||
1240 | /* | |
1241 | * Configurable global rate limit. | |
1242 | * | |
1243 | * ratelimit defines tokens/packet consumed for dst->rate_token | |
1244 | * bucket ratemask defines which icmp types are ratelimited by | |
1245 | * setting it's bit position. | |
1246 | * | |
1247 | * default: | |
1248 | * dest unreachable (3), source quench (4), | |
1249 | * time exceeded (11), parameter problem (12) | |
1250 | */ | |
1251 | ||
1252 | net->ipv4.sysctl_icmp_ratelimit = 1 * HZ; | |
1253 | net->ipv4.sysctl_icmp_ratemask = 0x1818; | |
1254 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0; | |
1255 | ||
a5710d65 DL |
1256 | return 0; |
1257 | ||
1258 | fail: | |
1d1c8d13 | 1259 | for_each_possible_cpu(i) |
349c9e3c ED |
1260 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i)); |
1261 | free_percpu(net->ipv4.icmp_sk); | |
a5710d65 | 1262 | return err; |
1da177e4 LT |
1263 | } |
1264 | ||
4a6ad7a1 DL |
1265 | static struct pernet_operations __net_initdata icmp_sk_ops = { |
1266 | .init = icmp_sk_init, | |
1267 | .exit = icmp_sk_exit, | |
1268 | }; | |
1269 | ||
1270 | int __init icmp_init(void) | |
1271 | { | |
959d2726 | 1272 | return register_pernet_subsys(&icmp_sk_ops); |
4a6ad7a1 | 1273 | } |