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 | |
257 | * without taking the spinlock. | |
258 | */ | |
259 | if (!icmp_global.credit) { | |
260 | delta = min_t(u32, now - icmp_global.stamp, HZ); | |
261 | if (delta < HZ / 50) | |
262 | return false; | |
263 | } | |
264 | ||
265 | spin_lock(&icmp_global.lock); | |
266 | delta = min_t(u32, now - icmp_global.stamp, HZ); | |
267 | if (delta >= HZ / 50) { | |
268 | incr = sysctl_icmp_msgs_per_sec * delta / HZ ; | |
269 | if (incr) | |
270 | icmp_global.stamp = now; | |
271 | } | |
272 | credit = min_t(u32, icmp_global.credit + incr, sysctl_icmp_msgs_burst); | |
273 | if (credit) { | |
274 | credit--; | |
275 | rc = true; | |
276 | } | |
277 | icmp_global.credit = credit; | |
278 | spin_unlock(&icmp_global.lock); | |
279 | return rc; | |
280 | } | |
281 | EXPORT_SYMBOL(icmp_global_allow); | |
282 | ||
c0303efe JDB |
283 | static bool icmpv4_mask_allow(struct net *net, int type, int code) |
284 | { | |
285 | if (type > NR_ICMP_TYPES) | |
286 | return true; | |
287 | ||
288 | /* Don't limit PMTU discovery. */ | |
289 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
290 | return true; | |
291 | ||
292 | /* Limit if icmp type is enabled in ratemask. */ | |
293 | if (!((1 << type) & net->ipv4.sysctl_icmp_ratemask)) | |
294 | return true; | |
295 | ||
296 | return false; | |
297 | } | |
298 | ||
299 | static bool icmpv4_global_allow(struct net *net, int type, int code) | |
300 | { | |
301 | if (icmpv4_mask_allow(net, type, code)) | |
302 | return true; | |
303 | ||
304 | if (icmp_global_allow()) | |
305 | return true; | |
306 | ||
307 | return false; | |
308 | } | |
309 | ||
1da177e4 LT |
310 | /* |
311 | * Send an ICMP frame. | |
312 | */ | |
313 | ||
4cdf507d ED |
314 | static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt, |
315 | struct flowi4 *fl4, int type, int code) | |
1da177e4 | 316 | { |
d8d1f30b | 317 | struct dst_entry *dst = &rt->dst; |
c0303efe | 318 | struct inet_peer *peer; |
92d86829 | 319 | bool rc = true; |
c0303efe | 320 | int vif; |
1da177e4 | 321 | |
c0303efe | 322 | if (icmpv4_mask_allow(net, type, code)) |
1da177e4 LT |
323 | goto out; |
324 | ||
325 | /* No rate limit on loopback */ | |
326 | if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) | |
e905a9ed | 327 | goto out; |
1da177e4 | 328 | |
c0303efe JDB |
329 | vif = l3mdev_master_ifindex(dst->dev); |
330 | peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, vif, 1); | |
331 | rc = inet_peer_xrlim_allow(peer, net->ipv4.sysctl_icmp_ratelimit); | |
332 | if (peer) | |
333 | inet_putpeer(peer); | |
1da177e4 LT |
334 | out: |
335 | return rc; | |
336 | } | |
337 | ||
338 | /* | |
339 | * Maintain the counters used in the SNMP statistics for outgoing ICMP | |
340 | */ | |
0388b004 | 341 | void icmp_out_count(struct net *net, unsigned char type) |
1da177e4 | 342 | { |
903fc196 | 343 | ICMPMSGOUT_INC_STATS(net, type); |
75c939bb | 344 | ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS); |
1da177e4 LT |
345 | } |
346 | ||
347 | /* | |
348 | * Checksum each fragment, and on the first include the headers and final | |
349 | * checksum. | |
350 | */ | |
351 | static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd, | |
352 | struct sk_buff *skb) | |
353 | { | |
354 | struct icmp_bxm *icmp_param = (struct icmp_bxm *)from; | |
5f92a738 | 355 | __wsum csum; |
1da177e4 LT |
356 | |
357 | csum = skb_copy_and_csum_bits(icmp_param->skb, | |
358 | icmp_param->offset + offset, | |
359 | to, len, 0); | |
360 | ||
361 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
362 | if (icmp_pointers[icmp_param->data.icmph.type].error) | |
363 | nf_ct_attach(skb, icmp_param->skb); | |
364 | return 0; | |
365 | } | |
366 | ||
367 | static void icmp_push_reply(struct icmp_bxm *icmp_param, | |
77968b78 | 368 | struct flowi4 *fl4, |
2e77d89b | 369 | struct ipcm_cookie *ipc, struct rtable **rt) |
1da177e4 | 370 | { |
1e3cf683 | 371 | struct sock *sk; |
1da177e4 LT |
372 | struct sk_buff *skb; |
373 | ||
d8d1f30b | 374 | sk = icmp_sk(dev_net((*rt)->dst.dev)); |
f5fca608 | 375 | if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param, |
e905a9ed YH |
376 | icmp_param->data_len+icmp_param->head_len, |
377 | icmp_param->head_len, | |
1f8438a8 | 378 | ipc, rt, MSG_DONTWAIT) < 0) { |
5d3848bc | 379 | __ICMP_INC_STATS(sock_net(sk), ICMP_MIB_OUTERRORS); |
1e3cf683 | 380 | ip_flush_pending_frames(sk); |
1f8438a8 | 381 | } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) { |
88c7664f | 382 | struct icmphdr *icmph = icmp_hdr(skb); |
d3bc23e7 | 383 | __wsum csum = 0; |
1da177e4 LT |
384 | struct sk_buff *skb1; |
385 | ||
1e3cf683 | 386 | skb_queue_walk(&sk->sk_write_queue, skb1) { |
1da177e4 LT |
387 | csum = csum_add(csum, skb1->csum); |
388 | } | |
389 | csum = csum_partial_copy_nocheck((void *)&icmp_param->data, | |
390 | (char *)icmph, | |
391 | icmp_param->head_len, csum); | |
392 | icmph->checksum = csum_fold(csum); | |
393 | skb->ip_summed = CHECKSUM_NONE; | |
77968b78 | 394 | ip_push_pending_frames(sk, fl4); |
1da177e4 LT |
395 | } |
396 | } | |
397 | ||
398 | /* | |
399 | * Driving logic for building and sending ICMP messages. | |
400 | */ | |
401 | ||
402 | static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb) | |
403 | { | |
1da177e4 | 404 | struct ipcm_cookie ipc; |
511c3f92 | 405 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 406 | struct net *net = dev_net(rt->dst.dev); |
77968b78 | 407 | struct flowi4 fl4; |
fdc0bde9 DL |
408 | struct sock *sk; |
409 | struct inet_sock *inet; | |
35ebf65e | 410 | __be32 daddr, saddr; |
e110861f | 411 | u32 mark = IP4_REPLY_MARK(net, skb->mark); |
c0303efe JDB |
412 | int type = icmp_param->data.icmph.type; |
413 | int code = icmp_param->data.icmph.code; | |
1da177e4 | 414 | |
91ed1e66 | 415 | if (ip_options_echo(net, &icmp_param->replyopts.opt.opt, skb)) |
f00c401b | 416 | return; |
1da177e4 | 417 | |
7ba91ecb JDB |
418 | /* Needed by both icmp_global_allow and icmp_xmit_lock */ |
419 | local_bh_disable(); | |
1da177e4 | 420 | |
c0303efe JDB |
421 | /* global icmp_msgs_per_sec */ |
422 | if (!icmpv4_global_allow(net, type, code)) | |
7ba91ecb JDB |
423 | goto out_bh_enable; |
424 | ||
425 | sk = icmp_xmit_lock(net); | |
426 | if (!sk) | |
427 | goto out_bh_enable; | |
428 | inet = inet_sk(sk); | |
c0303efe | 429 | |
1da177e4 | 430 | icmp_param->data.icmph.checksum = 0; |
1da177e4 | 431 | |
eddc9ec5 | 432 | inet->tos = ip_hdr(skb)->tos; |
e110861f | 433 | sk->sk_mark = mark; |
9f6abb5f | 434 | daddr = ipc.addr = ip_hdr(skb)->saddr; |
35ebf65e | 435 | saddr = fib_compute_spec_dst(skb); |
1da177e4 | 436 | ipc.opt = NULL; |
2244d07b | 437 | ipc.tx_flags = 0; |
aa661581 FF |
438 | ipc.ttl = 0; |
439 | ipc.tos = -1; | |
440 | ||
f6d8bd05 ED |
441 | if (icmp_param->replyopts.opt.opt.optlen) { |
442 | ipc.opt = &icmp_param->replyopts.opt; | |
443 | if (ipc.opt->opt.srr) | |
444 | daddr = icmp_param->replyopts.opt.opt.faddr; | |
1da177e4 | 445 | } |
77968b78 DM |
446 | memset(&fl4, 0, sizeof(fl4)); |
447 | fl4.daddr = daddr; | |
35ebf65e | 448 | fl4.saddr = saddr; |
e110861f | 449 | fl4.flowi4_mark = mark; |
e2d118a1 | 450 | fl4.flowi4_uid = sock_net_uid(net, NULL); |
77968b78 DM |
451 | fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos); |
452 | fl4.flowi4_proto = IPPROTO_ICMP; | |
385add90 | 453 | fl4.flowi4_oif = l3mdev_master_ifindex(skb->dev); |
77968b78 DM |
454 | security_skb_classify_flow(skb, flowi4_to_flowi(&fl4)); |
455 | rt = ip_route_output_key(net, &fl4); | |
456 | if (IS_ERR(rt)) | |
457 | goto out_unlock; | |
c0303efe | 458 | if (icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
77968b78 | 459 | icmp_push_reply(icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
460 | ip_rt_put(rt); |
461 | out_unlock: | |
405666db | 462 | icmp_xmit_unlock(sk); |
7ba91ecb JDB |
463 | out_bh_enable: |
464 | local_bh_enable(); | |
1da177e4 LT |
465 | } |
466 | ||
77968b78 DM |
467 | static struct rtable *icmp_route_lookup(struct net *net, |
468 | struct flowi4 *fl4, | |
469 | struct sk_buff *skb_in, | |
b71d1d42 | 470 | const struct iphdr *iph, |
e110861f | 471 | __be32 saddr, u8 tos, u32 mark, |
f6d460cf DM |
472 | int type, int code, |
473 | struct icmp_bxm *param) | |
474 | { | |
f6d460cf | 475 | struct rtable *rt, *rt2; |
415b3334 | 476 | struct flowi4 fl4_dec; |
f6d460cf DM |
477 | int err; |
478 | ||
77968b78 DM |
479 | memset(fl4, 0, sizeof(*fl4)); |
480 | fl4->daddr = (param->replyopts.opt.opt.srr ? | |
481 | param->replyopts.opt.opt.faddr : iph->saddr); | |
482 | fl4->saddr = saddr; | |
e110861f | 483 | fl4->flowi4_mark = mark; |
e2d118a1 | 484 | fl4->flowi4_uid = sock_net_uid(net, NULL); |
77968b78 DM |
485 | fl4->flowi4_tos = RT_TOS(tos); |
486 | fl4->flowi4_proto = IPPROTO_ICMP; | |
487 | fl4->fl4_icmp_type = type; | |
488 | fl4->fl4_icmp_code = code; | |
9d1a6c4e | 489 | fl4->flowi4_oif = l3mdev_master_ifindex(skb_dst(skb_in)->dev); |
613d09b3 | 490 | |
77968b78 | 491 | security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4)); |
3abd1ade | 492 | rt = ip_route_output_key_hash(net, fl4, skb_in); |
b23dd4fe DM |
493 | if (IS_ERR(rt)) |
494 | return rt; | |
f6d460cf DM |
495 | |
496 | /* No need to clone since we're just using its address. */ | |
497 | rt2 = rt; | |
498 | ||
9d6ec938 | 499 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
77968b78 | 500 | flowi4_to_flowi(fl4), NULL, 0); |
452edd59 | 501 | if (!IS_ERR(rt)) { |
f6d460cf DM |
502 | if (rt != rt2) |
503 | return rt; | |
452edd59 | 504 | } else if (PTR_ERR(rt) == -EPERM) { |
f6d460cf | 505 | rt = NULL; |
452edd59 DM |
506 | } else |
507 | return rt; | |
f6d460cf | 508 | |
415b3334 | 509 | err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET); |
f6d460cf DM |
510 | if (err) |
511 | goto relookup_failed; | |
512 | ||
9d1a6c4e | 513 | if (inet_addr_type_dev_table(net, skb_dst(skb_in)->dev, |
30bbaa19 | 514 | fl4_dec.saddr) == RTN_LOCAL) { |
415b3334 | 515 | rt2 = __ip_route_output_key(net, &fl4_dec); |
b23dd4fe DM |
516 | if (IS_ERR(rt2)) |
517 | err = PTR_ERR(rt2); | |
f6d460cf | 518 | } else { |
9d6ec938 | 519 | struct flowi4 fl4_2 = {}; |
f6d460cf DM |
520 | unsigned long orefdst; |
521 | ||
415b3334 | 522 | fl4_2.daddr = fl4_dec.saddr; |
9d6ec938 | 523 | rt2 = ip_route_output_key(net, &fl4_2); |
b23dd4fe DM |
524 | if (IS_ERR(rt2)) { |
525 | err = PTR_ERR(rt2); | |
f6d460cf | 526 | goto relookup_failed; |
b23dd4fe | 527 | } |
f6d460cf DM |
528 | /* Ugh! */ |
529 | orefdst = skb_in->_skb_refdst; /* save old refdst */ | |
773a69d6 | 530 | skb_dst_set(skb_in, NULL); |
415b3334 | 531 | err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr, |
f6d460cf DM |
532 | RT_TOS(tos), rt2->dst.dev); |
533 | ||
534 | dst_release(&rt2->dst); | |
535 | rt2 = skb_rtable(skb_in); | |
536 | skb_in->_skb_refdst = orefdst; /* restore old refdst */ | |
537 | } | |
538 | ||
539 | if (err) | |
540 | goto relookup_failed; | |
541 | ||
9d6ec938 | 542 | rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst, |
415b3334 | 543 | flowi4_to_flowi(&fl4_dec), NULL, |
9d6ec938 | 544 | XFRM_LOOKUP_ICMP); |
452edd59 | 545 | if (!IS_ERR(rt2)) { |
f6d460cf | 546 | dst_release(&rt->dst); |
415b3334 | 547 | memcpy(fl4, &fl4_dec, sizeof(*fl4)); |
f6d460cf | 548 | rt = rt2; |
452edd59 DM |
549 | } else if (PTR_ERR(rt2) == -EPERM) { |
550 | if (rt) | |
551 | dst_release(&rt->dst); | |
552 | return rt2; | |
553 | } else { | |
554 | err = PTR_ERR(rt2); | |
555 | goto relookup_failed; | |
f6d460cf | 556 | } |
f6d460cf DM |
557 | return rt; |
558 | ||
559 | relookup_failed: | |
560 | if (rt) | |
561 | return rt; | |
562 | return ERR_PTR(err); | |
563 | } | |
1da177e4 LT |
564 | |
565 | /* | |
566 | * Send an ICMP message in response to a situation | |
567 | * | |
568 | * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header. | |
569 | * MAY send more (we do). | |
570 | * MUST NOT change this header information. | |
571 | * MUST NOT reply to a multicast/broadcast IP address. | |
572 | * MUST NOT reply to a multicast/broadcast MAC address. | |
573 | * MUST reply to only the first fragment. | |
574 | */ | |
575 | ||
e4883014 | 576 | void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info) |
1da177e4 LT |
577 | { |
578 | struct iphdr *iph; | |
579 | int room; | |
8d9ba388 | 580 | struct icmp_bxm icmp_param; |
511c3f92 | 581 | struct rtable *rt = skb_rtable(skb_in); |
1da177e4 | 582 | struct ipcm_cookie ipc; |
77968b78 | 583 | struct flowi4 fl4; |
a61ced5d | 584 | __be32 saddr; |
1da177e4 | 585 | u8 tos; |
e110861f | 586 | u32 mark; |
dde1bc0e | 587 | struct net *net; |
4a6ad7a1 | 588 | struct sock *sk; |
1da177e4 LT |
589 | |
590 | if (!rt) | |
591 | goto out; | |
d8d1f30b | 592 | net = dev_net(rt->dst.dev); |
1da177e4 LT |
593 | |
594 | /* | |
595 | * Find the original header. It is expected to be valid, of course. | |
596 | * Check this, icmp_send is called from the most obscure devices | |
597 | * sometimes. | |
598 | */ | |
eddc9ec5 | 599 | iph = ip_hdr(skb_in); |
1da177e4 | 600 | |
27a884dc | 601 | if ((u8 *)iph < skb_in->head || |
f7c0c2ae SH |
602 | (skb_network_header(skb_in) + sizeof(*iph)) > |
603 | skb_tail_pointer(skb_in)) | |
1da177e4 LT |
604 | goto out; |
605 | ||
606 | /* | |
607 | * No replies to physical multicast/broadcast | |
608 | */ | |
609 | if (skb_in->pkt_type != PACKET_HOST) | |
610 | goto out; | |
611 | ||
612 | /* | |
613 | * Now check at the protocol level | |
614 | */ | |
615 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
616 | goto out; | |
617 | ||
618 | /* | |
619 | * Only reply to fragment 0. We byte re-order the constant | |
620 | * mask for efficiency. | |
621 | */ | |
622 | if (iph->frag_off & htons(IP_OFFSET)) | |
623 | goto out; | |
624 | ||
625 | /* | |
626 | * If we send an ICMP error to an ICMP error a mess would result.. | |
627 | */ | |
628 | if (icmp_pointers[type].error) { | |
629 | /* | |
630 | * We are an error, check if we are replying to an | |
631 | * ICMP error | |
632 | */ | |
633 | if (iph->protocol == IPPROTO_ICMP) { | |
634 | u8 _inner_type, *itp; | |
635 | ||
636 | itp = skb_header_pointer(skb_in, | |
d56f90a7 | 637 | skb_network_header(skb_in) + |
1da177e4 LT |
638 | (iph->ihl << 2) + |
639 | offsetof(struct icmphdr, | |
640 | type) - | |
641 | skb_in->data, | |
642 | sizeof(_inner_type), | |
643 | &_inner_type); | |
51456b29 | 644 | if (!itp) |
1da177e4 LT |
645 | goto out; |
646 | ||
647 | /* | |
648 | * Assume any unknown ICMP type is an error. This | |
649 | * isn't specified by the RFC, but think about it.. | |
650 | */ | |
651 | if (*itp > NR_ICMP_TYPES || | |
652 | icmp_pointers[*itp].error) | |
653 | goto out; | |
654 | } | |
655 | } | |
656 | ||
7ba91ecb JDB |
657 | /* Needed by both icmp_global_allow and icmp_xmit_lock */ |
658 | local_bh_disable(); | |
c0303efe | 659 | |
849a44de JDB |
660 | /* Check global sysctl_icmp_msgs_per_sec ratelimit, unless |
661 | * incoming dev is loopback. If outgoing dev change to not be | |
662 | * loopback, then peer ratelimit still work (in icmpv4_xrlim_allow) | |
663 | */ | |
664 | if (!(skb_in->dev && (skb_in->dev->flags&IFF_LOOPBACK)) && | |
665 | !icmpv4_global_allow(net, type, code)) | |
7ba91ecb JDB |
666 | goto out_bh_enable; |
667 | ||
668 | sk = icmp_xmit_lock(net); | |
669 | if (!sk) | |
670 | goto out_bh_enable; | |
1da177e4 LT |
671 | |
672 | /* | |
673 | * Construct source address and options. | |
674 | */ | |
675 | ||
676 | saddr = iph->daddr; | |
02a6d613 | 677 | if (!(rt->rt_flags & RTCF_LOCAL)) { |
6e1d9103 PM |
678 | struct net_device *dev = NULL; |
679 | ||
685c7944 | 680 | rcu_read_lock(); |
c7537967 DM |
681 | if (rt_is_input_route(rt) && |
682 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr) | |
92101b3b | 683 | dev = dev_get_by_index_rcu(net, inet_iif(skb_in)); |
6e1d9103 | 684 | |
685c7944 | 685 | if (dev) |
6e1d9103 | 686 | saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
685c7944 | 687 | else |
1c2fb7f9 | 688 | saddr = 0; |
685c7944 | 689 | rcu_read_unlock(); |
1c2fb7f9 | 690 | } |
1da177e4 LT |
691 | |
692 | tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) | | |
693 | IPTOS_PREC_INTERNETCONTROL) : | |
694 | iph->tos; | |
e110861f | 695 | mark = IP4_REPLY_MARK(net, skb_in->mark); |
1da177e4 | 696 | |
91ed1e66 | 697 | if (ip_options_echo(net, &icmp_param.replyopts.opt.opt, skb_in)) |
fa60cf7f | 698 | goto out_unlock; |
1da177e4 LT |
699 | |
700 | ||
701 | /* | |
702 | * Prepare data for ICMP header. | |
703 | */ | |
704 | ||
8d9ba388 JDB |
705 | icmp_param.data.icmph.type = type; |
706 | icmp_param.data.icmph.code = code; | |
707 | icmp_param.data.icmph.un.gateway = info; | |
708 | icmp_param.data.icmph.checksum = 0; | |
709 | icmp_param.skb = skb_in; | |
710 | icmp_param.offset = skb_network_offset(skb_in); | |
405666db | 711 | inet_sk(sk)->tos = tos; |
e110861f | 712 | sk->sk_mark = mark; |
1da177e4 | 713 | ipc.addr = iph->saddr; |
8d9ba388 | 714 | ipc.opt = &icmp_param.replyopts.opt; |
2244d07b | 715 | ipc.tx_flags = 0; |
aa661581 FF |
716 | ipc.ttl = 0; |
717 | ipc.tos = -1; | |
1da177e4 | 718 | |
e110861f | 719 | rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, mark, |
8d9ba388 | 720 | type, code, &icmp_param); |
f6d460cf DM |
721 | if (IS_ERR(rt)) |
722 | goto out_unlock; | |
1da177e4 | 723 | |
c0303efe | 724 | /* peer icmp_ratelimit */ |
a48eff12 | 725 | if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code)) |
1da177e4 LT |
726 | goto ende; |
727 | ||
728 | /* RFC says return as much as we can without exceeding 576 bytes. */ | |
729 | ||
d8d1f30b | 730 | room = dst_mtu(&rt->dst); |
1da177e4 LT |
731 | if (room > 576) |
732 | room = 576; | |
8d9ba388 | 733 | room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen; |
1da177e4 LT |
734 | room -= sizeof(struct icmphdr); |
735 | ||
8d9ba388 JDB |
736 | icmp_param.data_len = skb_in->len - icmp_param.offset; |
737 | if (icmp_param.data_len > room) | |
738 | icmp_param.data_len = room; | |
739 | icmp_param.head_len = sizeof(struct icmphdr); | |
1da177e4 | 740 | |
8d9ba388 | 741 | icmp_push_reply(&icmp_param, &fl4, &ipc, &rt); |
1da177e4 LT |
742 | ende: |
743 | ip_rt_put(rt); | |
744 | out_unlock: | |
405666db | 745 | icmp_xmit_unlock(sk); |
7ba91ecb JDB |
746 | out_bh_enable: |
747 | local_bh_enable(); | |
1da177e4 LT |
748 | out:; |
749 | } | |
4bc2f18b | 750 | EXPORT_SYMBOL(icmp_send); |
1da177e4 LT |
751 | |
752 | ||
1de9243b DM |
753 | static void icmp_socket_deliver(struct sk_buff *skb, u32 info) |
754 | { | |
755 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
756 | const struct net_protocol *ipprot; | |
757 | int protocol = iph->protocol; | |
758 | ||
f0a70e90 DM |
759 | /* Checkin full IP header plus 8 bytes of protocol to |
760 | * avoid additional coding at protocol handlers. | |
761 | */ | |
7304fe46 | 762 | if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) { |
5d3848bc | 763 | __ICMP_INC_STATS(dev_net(skb->dev), ICMP_MIB_INERRORS); |
f0a70e90 | 764 | return; |
7304fe46 | 765 | } |
f0a70e90 | 766 | |
1de9243b DM |
767 | raw_icmp_error(skb, protocol, info); |
768 | ||
1de9243b DM |
769 | ipprot = rcu_dereference(inet_protos[protocol]); |
770 | if (ipprot && ipprot->err_handler) | |
771 | ipprot->err_handler(skb, info); | |
1de9243b DM |
772 | } |
773 | ||
8ed1dc44 HFS |
774 | static bool icmp_tag_validation(int proto) |
775 | { | |
776 | bool ok; | |
777 | ||
778 | rcu_read_lock(); | |
779 | ok = rcu_dereference(inet_protos[proto])->icmp_strict_tag_validation; | |
780 | rcu_read_unlock(); | |
781 | return ok; | |
782 | } | |
783 | ||
1da177e4 | 784 | /* |
258bbb1b | 785 | * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEEDED, ICMP_QUENCH, and |
08578d8d | 786 | * ICMP_PARAMETERPROB. |
1da177e4 LT |
787 | */ |
788 | ||
e3e32170 | 789 | static bool icmp_unreach(struct sk_buff *skb) |
1da177e4 | 790 | { |
b71d1d42 | 791 | const struct iphdr *iph; |
1da177e4 | 792 | struct icmphdr *icmph; |
dde1bc0e | 793 | struct net *net; |
f9242b6b | 794 | u32 info = 0; |
dde1bc0e | 795 | |
adf30907 | 796 | net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
797 | |
798 | /* | |
799 | * Incomplete header ? | |
800 | * Only checks for the IP header, there should be an | |
801 | * additional check for longer headers in upper levels. | |
802 | */ | |
803 | ||
804 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
805 | goto out_err; | |
806 | ||
88c7664f | 807 | icmph = icmp_hdr(skb); |
b71d1d42 | 808 | iph = (const struct iphdr *)skb->data; |
1da177e4 LT |
809 | |
810 | if (iph->ihl < 5) /* Mangled header, drop. */ | |
811 | goto out_err; | |
812 | ||
258bbb1b MC |
813 | switch (icmph->type) { |
814 | case ICMP_DEST_UNREACH: | |
1da177e4 LT |
815 | switch (icmph->code & 15) { |
816 | case ICMP_NET_UNREACH: | |
817 | case ICMP_HOST_UNREACH: | |
818 | case ICMP_PROT_UNREACH: | |
819 | case ICMP_PORT_UNREACH: | |
820 | break; | |
821 | case ICMP_FRAG_NEEDED: | |
8ed1dc44 HFS |
822 | /* for documentation of the ip_no_pmtu_disc |
823 | * values please see | |
824 | * Documentation/networking/ip-sysctl.txt | |
825 | */ | |
826 | switch (net->ipv4.sysctl_ip_no_pmtu_disc) { | |
827 | default: | |
ba7a46f1 JP |
828 | net_dbg_ratelimited("%pI4: fragmentation needed and DF set\n", |
829 | &iph->daddr); | |
8ed1dc44 HFS |
830 | break; |
831 | case 2: | |
832 | goto out; | |
833 | case 3: | |
834 | if (!icmp_tag_validation(iph->protocol)) | |
835 | goto out; | |
836 | /* fall through */ | |
837 | case 0: | |
46517008 | 838 | info = ntohs(icmph->un.frag.mtu); |
1da177e4 LT |
839 | } |
840 | break; | |
841 | case ICMP_SR_FAILED: | |
ba7a46f1 JP |
842 | net_dbg_ratelimited("%pI4: Source Route Failed\n", |
843 | &iph->daddr); | |
1da177e4 LT |
844 | break; |
845 | default: | |
846 | break; | |
847 | } | |
848 | if (icmph->code > NR_ICMP_UNREACH) | |
849 | goto out; | |
258bbb1b MC |
850 | break; |
851 | case ICMP_PARAMETERPROB: | |
1da177e4 | 852 | info = ntohl(icmph->un.gateway) >> 24; |
258bbb1b MC |
853 | break; |
854 | case ICMP_TIME_EXCEEDED: | |
855 | __ICMP_INC_STATS(net, ICMP_MIB_INTIMEEXCDS); | |
856 | if (icmph->code == ICMP_EXC_FRAGTIME) | |
857 | goto out; | |
858 | break; | |
859 | } | |
1da177e4 LT |
860 | |
861 | /* | |
862 | * Throw it at our lower layers | |
863 | * | |
864 | * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed | |
865 | * header. | |
866 | * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the | |
867 | * transport layer. | |
868 | * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to | |
869 | * transport layer. | |
870 | */ | |
871 | ||
872 | /* | |
25985edc | 873 | * Check the other end isn't violating RFC 1122. Some routers send |
1da177e4 LT |
874 | * bogus responses to broadcast frames. If you see this message |
875 | * first check your netmask matches at both ends, if it does then | |
876 | * get the other vendor to fix their kit. | |
877 | */ | |
878 | ||
b34a95ee | 879 | if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses && |
30bbaa19 | 880 | inet_addr_type_dev_table(net, skb->dev, iph->daddr) == RTN_BROADCAST) { |
e87cc472 JP |
881 | net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n", |
882 | &ip_hdr(skb)->saddr, | |
883 | icmph->type, icmph->code, | |
884 | &iph->daddr, skb->dev->name); | |
1da177e4 LT |
885 | goto out; |
886 | } | |
887 | ||
1de9243b | 888 | icmp_socket_deliver(skb, info); |
1da177e4 LT |
889 | |
890 | out: | |
e3e32170 | 891 | return true; |
1da177e4 | 892 | out_err: |
5d3848bc | 893 | __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); |
e3e32170 | 894 | return false; |
1da177e4 LT |
895 | } |
896 | ||
897 | ||
898 | /* | |
899 | * Handle ICMP_REDIRECT. | |
900 | */ | |
901 | ||
e3e32170 | 902 | static bool icmp_redirect(struct sk_buff *skb) |
1da177e4 | 903 | { |
94206125 | 904 | if (skb->len < sizeof(struct iphdr)) { |
5d3848bc | 905 | __ICMP_INC_STATS(dev_net(skb->dev), ICMP_MIB_INERRORS); |
e3e32170 | 906 | return false; |
94206125 | 907 | } |
1da177e4 | 908 | |
e3e32170 RJ |
909 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) { |
910 | /* there aught to be a stat */ | |
911 | return false; | |
912 | } | |
c319b4d7 | 913 | |
d3351b75 | 914 | icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway); |
e3e32170 | 915 | return true; |
1da177e4 LT |
916 | } |
917 | ||
918 | /* | |
919 | * Handle ICMP_ECHO ("ping") requests. | |
920 | * | |
921 | * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo | |
922 | * requests. | |
923 | * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be | |
924 | * included in the reply. | |
925 | * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring | |
926 | * echo requests, MUST have default=NOT. | |
927 | * See also WRT handling of options once they are done and working. | |
928 | */ | |
929 | ||
e3e32170 | 930 | static bool icmp_echo(struct sk_buff *skb) |
1da177e4 | 931 | { |
b34a95ee PE |
932 | struct net *net; |
933 | ||
adf30907 | 934 | net = dev_net(skb_dst(skb)->dev); |
b34a95ee | 935 | if (!net->ipv4.sysctl_icmp_echo_ignore_all) { |
1da177e4 LT |
936 | struct icmp_bxm icmp_param; |
937 | ||
88c7664f | 938 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
939 | icmp_param.data.icmph.type = ICMP_ECHOREPLY; |
940 | icmp_param.skb = skb; | |
941 | icmp_param.offset = 0; | |
942 | icmp_param.data_len = skb->len; | |
943 | icmp_param.head_len = sizeof(struct icmphdr); | |
944 | icmp_reply(&icmp_param, skb); | |
945 | } | |
e3e32170 RJ |
946 | /* should there be an ICMP stat for ignored echos? */ |
947 | return true; | |
1da177e4 LT |
948 | } |
949 | ||
950 | /* | |
951 | * Handle ICMP Timestamp requests. | |
952 | * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests. | |
953 | * SHOULD be in the kernel for minimum random latency. | |
954 | * MUST be accurate to a few minutes. | |
955 | * MUST be updated at least at 15Hz. | |
956 | */ | |
e3e32170 | 957 | static bool icmp_timestamp(struct sk_buff *skb) |
1da177e4 | 958 | { |
1da177e4 LT |
959 | struct icmp_bxm icmp_param; |
960 | /* | |
961 | * Too short. | |
962 | */ | |
963 | if (skb->len < 4) | |
964 | goto out_err; | |
965 | ||
966 | /* | |
967 | * Fill in the current time as ms since midnight UT: | |
968 | */ | |
822c8685 | 969 | icmp_param.data.times[1] = inet_current_timestamp(); |
1da177e4 | 970 | icmp_param.data.times[2] = icmp_param.data.times[1]; |
15285402 GS |
971 | |
972 | BUG_ON(skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4)); | |
973 | ||
88c7664f | 974 | icmp_param.data.icmph = *icmp_hdr(skb); |
1da177e4 LT |
975 | icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY; |
976 | icmp_param.data.icmph.code = 0; | |
977 | icmp_param.skb = skb; | |
978 | icmp_param.offset = 0; | |
979 | icmp_param.data_len = 0; | |
980 | icmp_param.head_len = sizeof(struct icmphdr) + 12; | |
981 | icmp_reply(&icmp_param, skb); | |
e3e32170 RJ |
982 | return true; |
983 | ||
1da177e4 | 984 | out_err: |
5d3848bc | 985 | __ICMP_INC_STATS(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS); |
e3e32170 | 986 | return false; |
1da177e4 LT |
987 | } |
988 | ||
e3e32170 | 989 | static bool icmp_discard(struct sk_buff *skb) |
1da177e4 | 990 | { |
e3e32170 RJ |
991 | /* pretend it was a success */ |
992 | return true; | |
1da177e4 LT |
993 | } |
994 | ||
995 | /* | |
996 | * Deal with incoming ICMP packets. | |
997 | */ | |
998 | int icmp_rcv(struct sk_buff *skb) | |
999 | { | |
1000 | struct icmphdr *icmph; | |
511c3f92 | 1001 | struct rtable *rt = skb_rtable(skb); |
d8d1f30b | 1002 | struct net *net = dev_net(rt->dst.dev); |
e3e32170 | 1003 | bool success; |
1da177e4 | 1004 | |
aebcf82c | 1005 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { |
def8b4fa | 1006 | struct sec_path *sp = skb_sec_path(skb); |
8b7817f3 HX |
1007 | int nh; |
1008 | ||
def8b4fa | 1009 | if (!(sp && sp->xvec[sp->len - 1]->props.flags & |
aebcf82c HX |
1010 | XFRM_STATE_ICMP)) |
1011 | goto drop; | |
1012 | ||
8b7817f3 HX |
1013 | if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr))) |
1014 | goto drop; | |
1015 | ||
1016 | nh = skb_network_offset(skb); | |
1017 | skb_set_network_header(skb, sizeof(*icmph)); | |
1018 | ||
1019 | if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) | |
1020 | goto drop; | |
1021 | ||
1022 | skb_set_network_header(skb, nh); | |
1023 | } | |
1024 | ||
5d3848bc | 1025 | __ICMP_INC_STATS(net, ICMP_MIB_INMSGS); |
1da177e4 | 1026 | |
29a96e1f TH |
1027 | if (skb_checksum_simple_validate(skb)) |
1028 | goto csum_error; | |
1da177e4 | 1029 | |
8cf22943 HX |
1030 | if (!pskb_pull(skb, sizeof(*icmph))) |
1031 | goto error; | |
1da177e4 | 1032 | |
88c7664f | 1033 | icmph = icmp_hdr(skb); |
1da177e4 | 1034 | |
214d3f1f | 1035 | ICMPMSGIN_INC_STATS(net, icmph->type); |
1da177e4 LT |
1036 | /* |
1037 | * 18 is the highest 'known' ICMP type. Anything else is a mystery | |
1038 | * | |
1039 | * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently | |
1040 | * discarded. | |
1041 | */ | |
1042 | if (icmph->type > NR_ICMP_TYPES) | |
1043 | goto error; | |
1044 | ||
1045 | ||
1046 | /* | |
1047 | * Parse the ICMP message | |
1048 | */ | |
1049 | ||
e905a9ed | 1050 | if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
1da177e4 LT |
1051 | /* |
1052 | * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be | |
1053 | * silently ignored (we let user decide with a sysctl). | |
1054 | * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently | |
1055 | * discarded if to broadcast/multicast. | |
1056 | */ | |
4c866aa7 AK |
1057 | if ((icmph->type == ICMP_ECHO || |
1058 | icmph->type == ICMP_TIMESTAMP) && | |
b34a95ee | 1059 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts) { |
1da177e4 LT |
1060 | goto error; |
1061 | } | |
1062 | if (icmph->type != ICMP_ECHO && | |
1063 | icmph->type != ICMP_TIMESTAMP && | |
1064 | icmph->type != ICMP_ADDRESS && | |
1065 | icmph->type != ICMP_ADDRESSREPLY) { | |
1066 | goto error; | |
e905a9ed | 1067 | } |
1da177e4 LT |
1068 | } |
1069 | ||
e3e32170 RJ |
1070 | success = icmp_pointers[icmph->type].handler(skb); |
1071 | ||
1072 | if (success) { | |
1073 | consume_skb(skb); | |
f91c58d6 | 1074 | return NET_RX_SUCCESS; |
e3e32170 | 1075 | } |
1da177e4 LT |
1076 | |
1077 | drop: | |
1078 | kfree_skb(skb); | |
f91c58d6 | 1079 | return NET_RX_DROP; |
6a5dc9e5 | 1080 | csum_error: |
5d3848bc | 1081 | __ICMP_INC_STATS(net, ICMP_MIB_CSUMERRORS); |
1da177e4 | 1082 | error: |
5d3848bc | 1083 | __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
1084 | goto drop; |
1085 | } | |
1086 | ||
5b052042 LW |
1087 | void icmp_err(struct sk_buff *skb, u32 info) |
1088 | { | |
1089 | struct iphdr *iph = (struct iphdr *)skb->data; | |
6d0bfe22 LC |
1090 | int offset = iph->ihl<<2; |
1091 | struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset); | |
5b052042 LW |
1092 | int type = icmp_hdr(skb)->type; |
1093 | int code = icmp_hdr(skb)->code; | |
1094 | struct net *net = dev_net(skb->dev); | |
1095 | ||
1096 | /* | |
1097 | * Use ping_err to handle all icmp errors except those | |
1098 | * triggered by ICMP_ECHOREPLY which sent from kernel. | |
1099 | */ | |
1100 | if (icmph->type != ICMP_ECHOREPLY) { | |
6d0bfe22 | 1101 | ping_err(skb, offset, info); |
5b052042 LW |
1102 | return; |
1103 | } | |
1104 | ||
1105 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) | |
1106 | ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0); | |
1107 | else if (type == ICMP_REDIRECT) | |
1108 | ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0); | |
1109 | } | |
1110 | ||
1da177e4 LT |
1111 | /* |
1112 | * This table is the definition of how we handle ICMP. | |
1113 | */ | |
9b5b5cff | 1114 | static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = { |
1da177e4 | 1115 | [ICMP_ECHOREPLY] = { |
c319b4d7 | 1116 | .handler = ping_rcv, |
1da177e4 LT |
1117 | }, |
1118 | [1] = { | |
1da177e4 LT |
1119 | .handler = icmp_discard, |
1120 | .error = 1, | |
1121 | }, | |
1122 | [2] = { | |
1da177e4 LT |
1123 | .handler = icmp_discard, |
1124 | .error = 1, | |
1125 | }, | |
1126 | [ICMP_DEST_UNREACH] = { | |
1da177e4 LT |
1127 | .handler = icmp_unreach, |
1128 | .error = 1, | |
1129 | }, | |
1130 | [ICMP_SOURCE_QUENCH] = { | |
1da177e4 LT |
1131 | .handler = icmp_unreach, |
1132 | .error = 1, | |
1133 | }, | |
1134 | [ICMP_REDIRECT] = { | |
1da177e4 LT |
1135 | .handler = icmp_redirect, |
1136 | .error = 1, | |
1137 | }, | |
1138 | [6] = { | |
1da177e4 LT |
1139 | .handler = icmp_discard, |
1140 | .error = 1, | |
1141 | }, | |
1142 | [7] = { | |
1da177e4 LT |
1143 | .handler = icmp_discard, |
1144 | .error = 1, | |
1145 | }, | |
1146 | [ICMP_ECHO] = { | |
1da177e4 LT |
1147 | .handler = icmp_echo, |
1148 | }, | |
1149 | [9] = { | |
1da177e4 LT |
1150 | .handler = icmp_discard, |
1151 | .error = 1, | |
1152 | }, | |
1153 | [10] = { | |
1da177e4 LT |
1154 | .handler = icmp_discard, |
1155 | .error = 1, | |
1156 | }, | |
1157 | [ICMP_TIME_EXCEEDED] = { | |
1da177e4 LT |
1158 | .handler = icmp_unreach, |
1159 | .error = 1, | |
1160 | }, | |
1161 | [ICMP_PARAMETERPROB] = { | |
1da177e4 LT |
1162 | .handler = icmp_unreach, |
1163 | .error = 1, | |
1164 | }, | |
1165 | [ICMP_TIMESTAMP] = { | |
1da177e4 LT |
1166 | .handler = icmp_timestamp, |
1167 | }, | |
1168 | [ICMP_TIMESTAMPREPLY] = { | |
1da177e4 LT |
1169 | .handler = icmp_discard, |
1170 | }, | |
1171 | [ICMP_INFO_REQUEST] = { | |
1da177e4 LT |
1172 | .handler = icmp_discard, |
1173 | }, | |
e905a9ed | 1174 | [ICMP_INFO_REPLY] = { |
1da177e4 LT |
1175 | .handler = icmp_discard, |
1176 | }, | |
1177 | [ICMP_ADDRESS] = { | |
838942a5 | 1178 | .handler = icmp_discard, |
1da177e4 LT |
1179 | }, |
1180 | [ICMP_ADDRESSREPLY] = { | |
838942a5 | 1181 | .handler = icmp_discard, |
1da177e4 LT |
1182 | }, |
1183 | }; | |
1184 | ||
4a6ad7a1 | 1185 | static void __net_exit icmp_sk_exit(struct net *net) |
1da177e4 | 1186 | { |
1da177e4 LT |
1187 | int i; |
1188 | ||
5c8cafd6 | 1189 | for_each_possible_cpu(i) |
349c9e3c ED |
1190 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i)); |
1191 | free_percpu(net->ipv4.icmp_sk); | |
4a6ad7a1 | 1192 | net->ipv4.icmp_sk = NULL; |
a5710d65 | 1193 | } |
1da177e4 | 1194 | |
263173af | 1195 | static int __net_init icmp_sk_init(struct net *net) |
a5710d65 | 1196 | { |
a5710d65 DL |
1197 | int i, err; |
1198 | ||
349c9e3c ED |
1199 | net->ipv4.icmp_sk = alloc_percpu(struct sock *); |
1200 | if (!net->ipv4.icmp_sk) | |
79c91159 DL |
1201 | return -ENOMEM; |
1202 | ||
a5710d65 | 1203 | for_each_possible_cpu(i) { |
1e3cf683 | 1204 | struct sock *sk; |
1da177e4 | 1205 | |
c1e9894d DL |
1206 | err = inet_ctl_sock_create(&sk, PF_INET, |
1207 | SOCK_RAW, IPPROTO_ICMP, net); | |
1da177e4 | 1208 | if (err < 0) |
a5710d65 | 1209 | goto fail; |
1da177e4 | 1210 | |
349c9e3c | 1211 | *per_cpu_ptr(net->ipv4.icmp_sk, i) = sk; |
1da177e4 LT |
1212 | |
1213 | /* Enough space for 2 64K ICMP packets, including | |
87fb4b7b | 1214 | * sk_buff/skb_shared_info struct overhead. |
1da177e4 | 1215 | */ |
87fb4b7b | 1216 | sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); |
1da177e4 | 1217 | |
b3a5b6cc ED |
1218 | /* |
1219 | * Speedup sock_wfree() | |
1220 | */ | |
1221 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
c1e9894d | 1222 | inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT; |
1da177e4 | 1223 | } |
a24022e1 PE |
1224 | |
1225 | /* Control parameters for ECHO replies. */ | |
1226 | net->ipv4.sysctl_icmp_echo_ignore_all = 0; | |
1227 | net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1; | |
1228 | ||
1229 | /* Control parameter - ignore bogus broadcast responses? */ | |
1230 | net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1; | |
1231 | ||
1232 | /* | |
1233 | * Configurable global rate limit. | |
1234 | * | |
1235 | * ratelimit defines tokens/packet consumed for dst->rate_token | |
1236 | * bucket ratemask defines which icmp types are ratelimited by | |
1237 | * setting it's bit position. | |
1238 | * | |
1239 | * default: | |
1240 | * dest unreachable (3), source quench (4), | |
1241 | * time exceeded (11), parameter problem (12) | |
1242 | */ | |
1243 | ||
1244 | net->ipv4.sysctl_icmp_ratelimit = 1 * HZ; | |
1245 | net->ipv4.sysctl_icmp_ratemask = 0x1818; | |
1246 | net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0; | |
1247 | ||
a5710d65 DL |
1248 | return 0; |
1249 | ||
1250 | fail: | |
1d1c8d13 | 1251 | for_each_possible_cpu(i) |
349c9e3c ED |
1252 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.icmp_sk, i)); |
1253 | free_percpu(net->ipv4.icmp_sk); | |
a5710d65 | 1254 | return err; |
1da177e4 LT |
1255 | } |
1256 | ||
4a6ad7a1 DL |
1257 | static struct pernet_operations __net_initdata icmp_sk_ops = { |
1258 | .init = icmp_sk_init, | |
1259 | .exit = icmp_sk_exit, | |
1260 | }; | |
1261 | ||
1262 | int __init icmp_init(void) | |
1263 | { | |
959d2726 | 1264 | return register_pernet_subsys(&icmp_sk_ops); |
4a6ad7a1 | 1265 | } |