drivers: power: report battery voltage in AOSP compatible format
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv4 / icmp.c
CommitLineData
1da177e4
LT
1/*
2 * NET3: Implementation of the ICMP protocol layer.
3 *
113aa838 4 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4 5 *
1da177e4
LT
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 * Some of the function names and the icmp unreach table for this
12 * module were derived from [icmp.c 1.0.11 06/02/93] by
13 * Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
14 * Other than that this module is a complete rewrite.
15 *
16 * Fixes:
17 * Clemens Fruhwirth : introduce global icmp rate limiting
18 * with icmp type masking ability instead
19 * of broken per type icmp timeouts.
20 * Mike Shaver : RFC1122 checks.
21 * Alan Cox : Multicast ping reply as self.
22 * Alan Cox : Fix atomicity lockup in ip_build_xmit
23 * call.
24 * Alan Cox : Added 216,128 byte paths to the MTU
25 * code.
26 * Martin Mares : RFC1812 checks.
27 * Martin Mares : Can be configured to follow redirects
28 * if acting as a router _without_ a
29 * routing protocol (RFC 1812).
30 * Martin Mares : Echo requests may be configured to
31 * be ignored (RFC 1812).
32 * Martin Mares : Limitation of ICMP error message
33 * transmit rate (RFC 1812).
34 * Martin Mares : TOS and Precedence set correctly
35 * (RFC 1812).
36 * Martin Mares : Now copying as much data from the
37 * original packet as we can without
38 * exceeding 576 bytes (RFC 1812).
39 * Willy Konynenberg : Transparent proxying support.
40 * Keith Owens : RFC1191 correction for 4.2BSD based
41 * path MTU bug.
42 * Thomas Quinot : ICMP Dest Unreach codes up to 15 are
43 * valid (RFC 1812).
44 * Andi Kleen : Check all packet lengths properly
45 * and moved all kfree_skb() up to
46 * icmp_rcv.
47 * Andi Kleen : Move the rate limit bookkeeping
48 * into the dest entry and use a token
49 * bucket filter (thanks to ANK). Make
50 * the rates sysctl configurable.
51 * Yu Tianli : Fixed two ugly bugs in icmp_send
52 * - IP option length was accounted wrongly
53 * - ICMP header length was not accounted
54 * at all.
55 * Tristan Greaves : Added sysctl option to ignore bogus
56 * broadcast responses from broken routers.
57 *
58 * To Fix:
59 *
60 * - Should use skb_pull() instead of all the manual checking.
61 * This would also greatly simply some upper layer error handlers. --AK
62 *
63 */
64
afd46503
JP
65#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
66
1da177e4
LT
67#include <linux/module.h>
68#include <linux/types.h>
69#include <linux/jiffies.h>
70#include <linux/kernel.h>
71#include <linux/fcntl.h>
72#include <linux/socket.h>
73#include <linux/in.h>
74#include <linux/inet.h>
14c85021 75#include <linux/inetdevice.h>
1da177e4
LT
76#include <linux/netdevice.h>
77#include <linux/string.h>
78#include <linux/netfilter_ipv4.h>
5a0e3ad6 79#include <linux/slab.h>
1da177e4
LT
80#include <net/snmp.h>
81#include <net/ip.h>
82#include <net/route.h>
83#include <net/protocol.h>
84#include <net/icmp.h>
85#include <net/tcp.h>
86#include <net/udp.h>
87#include <net/raw.h>
c319b4d7 88#include <net/ping.h>
1da177e4
LT
89#include <linux/skbuff.h>
90#include <net/sock.h>
91#include <linux/errno.h>
92#include <linux/timer.h>
93#include <linux/init.h>
1da177e4
LT
94#include <asm/uaccess.h>
95#include <net/checksum.h>
8b7817f3 96#include <net/xfrm.h>
c1e9894d 97#include <net/inet_common.h>
35ebf65e 98#include <net/ip_fib.h>
1da177e4
LT
99
100/*
101 * Build xmit assembly blocks
102 */
103
104struct icmp_bxm {
105 struct sk_buff *skb;
106 int offset;
107 int data_len;
108
109 struct {
110 struct icmphdr icmph;
b03d73e3 111 __be32 times[3];
1da177e4
LT
112 } data;
113 int head_len;
f6d8bd05 114 struct ip_options_data replyopts;
1da177e4
LT
115};
116
1da177e4
LT
117/* An array of errno for error messages from dest unreach. */
118/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
119
e754834e 120const struct icmp_err icmp_err_convert[] = {
1da177e4
LT
121 {
122 .errno = ENETUNREACH, /* ICMP_NET_UNREACH */
123 .fatal = 0,
124 },
125 {
126 .errno = EHOSTUNREACH, /* ICMP_HOST_UNREACH */
127 .fatal = 0,
128 },
129 {
130 .errno = ENOPROTOOPT /* ICMP_PROT_UNREACH */,
131 .fatal = 1,
132 },
133 {
134 .errno = ECONNREFUSED, /* ICMP_PORT_UNREACH */
135 .fatal = 1,
136 },
137 {
138 .errno = EMSGSIZE, /* ICMP_FRAG_NEEDED */
139 .fatal = 0,
140 },
141 {
142 .errno = EOPNOTSUPP, /* ICMP_SR_FAILED */
143 .fatal = 0,
144 },
145 {
146 .errno = ENETUNREACH, /* ICMP_NET_UNKNOWN */
147 .fatal = 1,
148 },
149 {
150 .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
151 .fatal = 1,
152 },
153 {
154 .errno = ENONET, /* ICMP_HOST_ISOLATED */
155 .fatal = 1,
156 },
157 {
158 .errno = ENETUNREACH, /* ICMP_NET_ANO */
159 .fatal = 1,
160 },
161 {
162 .errno = EHOSTUNREACH, /* ICMP_HOST_ANO */
163 .fatal = 1,
164 },
165 {
166 .errno = ENETUNREACH, /* ICMP_NET_UNR_TOS */
167 .fatal = 0,
168 },
169 {
170 .errno = EHOSTUNREACH, /* ICMP_HOST_UNR_TOS */
171 .fatal = 0,
172 },
173 {
174 .errno = EHOSTUNREACH, /* ICMP_PKT_FILTERED */
175 .fatal = 1,
176 },
177 {
178 .errno = EHOSTUNREACH, /* ICMP_PREC_VIOLATION */
179 .fatal = 1,
180 },
181 {
182 .errno = EHOSTUNREACH, /* ICMP_PREC_CUTOFF */
183 .fatal = 1,
184 },
185};
4bc2f18b 186EXPORT_SYMBOL(icmp_err_convert);
1da177e4 187
1da177e4
LT
188/*
189 * ICMP control array. This specifies what to do with each ICMP.
190 */
191
192struct icmp_control {
1da177e4
LT
193 void (*handler)(struct sk_buff *skb);
194 short error; /* This ICMP is classed as an error message */
195};
196
9b5b5cff 197static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
1da177e4
LT
198
199/*
200 * The ICMP socket(s). This is the most convenient way to flow control
201 * our ICMP output as well as maintain a clean interface throughout
202 * all layers. All Socketless IP sends will soon be gone.
203 *
204 * On SMP we have one ICMP socket per-cpu.
205 */
4a6ad7a1
DL
206static struct sock *icmp_sk(struct net *net)
207{
208 return net->ipv4.icmp_sk[smp_processor_id()];
209}
1da177e4 210
fdc0bde9 211static inline struct sock *icmp_xmit_lock(struct net *net)
1da177e4 212{
fdc0bde9
DL
213 struct sock *sk;
214
1da177e4
LT
215 local_bh_disable();
216
fdc0bde9
DL
217 sk = icmp_sk(net);
218
405666db 219 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
1da177e4
LT
220 /* This can happen if the output path signals a
221 * dst_link_failure() for an outgoing ICMP packet.
222 */
223 local_bh_enable();
fdc0bde9 224 return NULL;
1da177e4 225 }
fdc0bde9 226 return sk;
1da177e4
LT
227}
228
405666db 229static inline void icmp_xmit_unlock(struct sock *sk)
1da177e4 230{
405666db 231 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT
232}
233
234/*
235 * Send an ICMP frame.
236 */
237
92d86829 238static inline bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
a48eff12 239 struct flowi4 *fl4, int type, int code)
1da177e4 240{
d8d1f30b 241 struct dst_entry *dst = &rt->dst;
92d86829 242 bool rc = true;
1da177e4
LT
243
244 if (type > NR_ICMP_TYPES)
245 goto out;
246
247 /* Don't limit PMTU discovery. */
248 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
249 goto out;
250
251 /* No rate limit on loopback */
252 if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
e905a9ed 253 goto out;
1da177e4
LT
254
255 /* Limit if icmp type is enabled in ratemask. */
92d86829 256 if ((1 << type) & net->ipv4.sysctl_icmp_ratemask) {
1d861aa4 257 struct inet_peer *peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, 1);
fbfe95a4 258 rc = inet_peer_xrlim_allow(peer,
92d86829 259 net->ipv4.sysctl_icmp_ratelimit);
e1a67642
NC
260 if (peer)
261 inet_putpeer(peer);
92d86829 262 }
1da177e4
LT
263out:
264 return rc;
265}
266
267/*
268 * Maintain the counters used in the SNMP statistics for outgoing ICMP
269 */
0388b004 270void icmp_out_count(struct net *net, unsigned char type)
1da177e4 271{
903fc196 272 ICMPMSGOUT_INC_STATS(net, type);
75c939bb 273 ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
1da177e4
LT
274}
275
276/*
277 * Checksum each fragment, and on the first include the headers and final
278 * checksum.
279 */
280static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
281 struct sk_buff *skb)
282{
283 struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
5f92a738 284 __wsum csum;
1da177e4
LT
285
286 csum = skb_copy_and_csum_bits(icmp_param->skb,
287 icmp_param->offset + offset,
288 to, len, 0);
289
290 skb->csum = csum_block_add(skb->csum, csum, odd);
291 if (icmp_pointers[icmp_param->data.icmph.type].error)
292 nf_ct_attach(skb, icmp_param->skb);
293 return 0;
294}
295
296static void icmp_push_reply(struct icmp_bxm *icmp_param,
77968b78 297 struct flowi4 *fl4,
2e77d89b 298 struct ipcm_cookie *ipc, struct rtable **rt)
1da177e4 299{
1e3cf683 300 struct sock *sk;
1da177e4
LT
301 struct sk_buff *skb;
302
d8d1f30b 303 sk = icmp_sk(dev_net((*rt)->dst.dev));
f5fca608 304 if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
e905a9ed
YH
305 icmp_param->data_len+icmp_param->head_len,
306 icmp_param->head_len,
1f8438a8
ED
307 ipc, rt, MSG_DONTWAIT) < 0) {
308 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_OUTERRORS);
1e3cf683 309 ip_flush_pending_frames(sk);
1f8438a8 310 } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
88c7664f 311 struct icmphdr *icmph = icmp_hdr(skb);
d3bc23e7 312 __wsum csum = 0;
1da177e4
LT
313 struct sk_buff *skb1;
314
1e3cf683 315 skb_queue_walk(&sk->sk_write_queue, skb1) {
1da177e4
LT
316 csum = csum_add(csum, skb1->csum);
317 }
318 csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
319 (char *)icmph,
320 icmp_param->head_len, csum);
321 icmph->checksum = csum_fold(csum);
322 skb->ip_summed = CHECKSUM_NONE;
77968b78 323 ip_push_pending_frames(sk, fl4);
1da177e4
LT
324 }
325}
326
327/*
328 * Driving logic for building and sending ICMP messages.
329 */
330
331static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
332{
1da177e4 333 struct ipcm_cookie ipc;
511c3f92 334 struct rtable *rt = skb_rtable(skb);
d8d1f30b 335 struct net *net = dev_net(rt->dst.dev);
77968b78 336 struct flowi4 fl4;
fdc0bde9
DL
337 struct sock *sk;
338 struct inet_sock *inet;
35ebf65e 339 __be32 daddr, saddr;
6fa3eb70 340 u32 mark = IP4_REPLY_MARK(net, skb->mark);
1da177e4 341
f6d8bd05 342 if (ip_options_echo(&icmp_param->replyopts.opt.opt, skb))
f00c401b 343 return;
1da177e4 344
fdc0bde9
DL
345 sk = icmp_xmit_lock(net);
346 if (sk == NULL)
1da177e4 347 return;
fdc0bde9 348 inet = inet_sk(sk);
1da177e4
LT
349
350 icmp_param->data.icmph.checksum = 0;
1da177e4 351
eddc9ec5 352 inet->tos = ip_hdr(skb)->tos;
6fa3eb70 353 sk->sk_mark = mark;
9f6abb5f 354 daddr = ipc.addr = ip_hdr(skb)->saddr;
35ebf65e 355 saddr = fib_compute_spec_dst(skb);
1da177e4 356 ipc.opt = NULL;
2244d07b 357 ipc.tx_flags = 0;
f6d8bd05
ED
358 if (icmp_param->replyopts.opt.opt.optlen) {
359 ipc.opt = &icmp_param->replyopts.opt;
360 if (ipc.opt->opt.srr)
361 daddr = icmp_param->replyopts.opt.opt.faddr;
1da177e4 362 }
77968b78
DM
363 memset(&fl4, 0, sizeof(fl4));
364 fl4.daddr = daddr;
35ebf65e 365 fl4.saddr = saddr;
6fa3eb70 366 fl4.flowi4_mark = mark;
77968b78
DM
367 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
368 fl4.flowi4_proto = IPPROTO_ICMP;
369 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
370 rt = ip_route_output_key(net, &fl4);
371 if (IS_ERR(rt))
372 goto out_unlock;
a48eff12 373 if (icmpv4_xrlim_allow(net, rt, &fl4, icmp_param->data.icmph.type,
1da177e4 374 icmp_param->data.icmph.code))
77968b78 375 icmp_push_reply(icmp_param, &fl4, &ipc, &rt);
1da177e4
LT
376 ip_rt_put(rt);
377out_unlock:
405666db 378 icmp_xmit_unlock(sk);
1da177e4
LT
379}
380
77968b78
DM
381static struct rtable *icmp_route_lookup(struct net *net,
382 struct flowi4 *fl4,
383 struct sk_buff *skb_in,
b71d1d42 384 const struct iphdr *iph,
6fa3eb70 385 __be32 saddr, u8 tos, u32 mark,
f6d460cf
DM
386 int type, int code,
387 struct icmp_bxm *param)
388{
f6d460cf 389 struct rtable *rt, *rt2;
415b3334 390 struct flowi4 fl4_dec;
f6d460cf
DM
391 int err;
392
77968b78
DM
393 memset(fl4, 0, sizeof(*fl4));
394 fl4->daddr = (param->replyopts.opt.opt.srr ?
395 param->replyopts.opt.opt.faddr : iph->saddr);
396 fl4->saddr = saddr;
6fa3eb70 397 fl4->flowi4_mark = mark;
77968b78
DM
398 fl4->flowi4_tos = RT_TOS(tos);
399 fl4->flowi4_proto = IPPROTO_ICMP;
400 fl4->fl4_icmp_type = type;
401 fl4->fl4_icmp_code = code;
402 security_skb_classify_flow(skb_in, flowi4_to_flowi(fl4));
403 rt = __ip_route_output_key(net, fl4);
b23dd4fe
DM
404 if (IS_ERR(rt))
405 return rt;
f6d460cf
DM
406
407 /* No need to clone since we're just using its address. */
408 rt2 = rt;
409
9d6ec938 410 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
77968b78 411 flowi4_to_flowi(fl4), NULL, 0);
452edd59 412 if (!IS_ERR(rt)) {
f6d460cf
DM
413 if (rt != rt2)
414 return rt;
452edd59 415 } else if (PTR_ERR(rt) == -EPERM) {
f6d460cf 416 rt = NULL;
452edd59
DM
417 } else
418 return rt;
f6d460cf 419
415b3334 420 err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
f6d460cf
DM
421 if (err)
422 goto relookup_failed;
423
415b3334
DM
424 if (inet_addr_type(net, fl4_dec.saddr) == RTN_LOCAL) {
425 rt2 = __ip_route_output_key(net, &fl4_dec);
b23dd4fe
DM
426 if (IS_ERR(rt2))
427 err = PTR_ERR(rt2);
f6d460cf 428 } else {
9d6ec938 429 struct flowi4 fl4_2 = {};
f6d460cf
DM
430 unsigned long orefdst;
431
415b3334 432 fl4_2.daddr = fl4_dec.saddr;
9d6ec938 433 rt2 = ip_route_output_key(net, &fl4_2);
b23dd4fe
DM
434 if (IS_ERR(rt2)) {
435 err = PTR_ERR(rt2);
f6d460cf 436 goto relookup_failed;
b23dd4fe 437 }
f6d460cf
DM
438 /* Ugh! */
439 orefdst = skb_in->_skb_refdst; /* save old refdst */
415b3334 440 err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
f6d460cf
DM
441 RT_TOS(tos), rt2->dst.dev);
442
443 dst_release(&rt2->dst);
444 rt2 = skb_rtable(skb_in);
445 skb_in->_skb_refdst = orefdst; /* restore old refdst */
446 }
447
448 if (err)
449 goto relookup_failed;
450
9d6ec938 451 rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
415b3334 452 flowi4_to_flowi(&fl4_dec), NULL,
9d6ec938 453 XFRM_LOOKUP_ICMP);
452edd59 454 if (!IS_ERR(rt2)) {
f6d460cf 455 dst_release(&rt->dst);
415b3334 456 memcpy(fl4, &fl4_dec, sizeof(*fl4));
f6d460cf 457 rt = rt2;
452edd59
DM
458 } else if (PTR_ERR(rt2) == -EPERM) {
459 if (rt)
460 dst_release(&rt->dst);
461 return rt2;
462 } else {
463 err = PTR_ERR(rt2);
464 goto relookup_failed;
f6d460cf 465 }
f6d460cf
DM
466 return rt;
467
468relookup_failed:
469 if (rt)
470 return rt;
471 return ERR_PTR(err);
472}
1da177e4
LT
473
474/*
475 * Send an ICMP message in response to a situation
476 *
477 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
478 * MAY send more (we do).
479 * MUST NOT change this header information.
480 * MUST NOT reply to a multicast/broadcast IP address.
481 * MUST NOT reply to a multicast/broadcast MAC address.
482 * MUST reply to only the first fragment.
483 */
484
e4883014 485void icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info)
1da177e4
LT
486{
487 struct iphdr *iph;
488 int room;
489 struct icmp_bxm icmp_param;
511c3f92 490 struct rtable *rt = skb_rtable(skb_in);
1da177e4 491 struct ipcm_cookie ipc;
77968b78 492 struct flowi4 fl4;
a61ced5d 493 __be32 saddr;
1da177e4 494 u8 tos;
6fa3eb70 495 u32 mark;
dde1bc0e 496 struct net *net;
4a6ad7a1 497 struct sock *sk;
1da177e4
LT
498
499 if (!rt)
500 goto out;
d8d1f30b 501 net = dev_net(rt->dst.dev);
1da177e4
LT
502
503 /*
504 * Find the original header. It is expected to be valid, of course.
505 * Check this, icmp_send is called from the most obscure devices
506 * sometimes.
507 */
eddc9ec5 508 iph = ip_hdr(skb_in);
1da177e4 509
27a884dc
ACM
510 if ((u8 *)iph < skb_in->head ||
511 (skb_in->network_header + sizeof(*iph)) > skb_in->tail)
1da177e4
LT
512 goto out;
513
514 /*
515 * No replies to physical multicast/broadcast
516 */
517 if (skb_in->pkt_type != PACKET_HOST)
518 goto out;
519
520 /*
521 * Now check at the protocol level
522 */
523 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
524 goto out;
525
526 /*
527 * Only reply to fragment 0. We byte re-order the constant
528 * mask for efficiency.
529 */
530 if (iph->frag_off & htons(IP_OFFSET))
531 goto out;
532
533 /*
534 * If we send an ICMP error to an ICMP error a mess would result..
535 */
536 if (icmp_pointers[type].error) {
537 /*
538 * We are an error, check if we are replying to an
539 * ICMP error
540 */
541 if (iph->protocol == IPPROTO_ICMP) {
542 u8 _inner_type, *itp;
543
544 itp = skb_header_pointer(skb_in,
d56f90a7 545 skb_network_header(skb_in) +
1da177e4
LT
546 (iph->ihl << 2) +
547 offsetof(struct icmphdr,
548 type) -
549 skb_in->data,
550 sizeof(_inner_type),
551 &_inner_type);
552 if (itp == NULL)
553 goto out;
554
555 /*
556 * Assume any unknown ICMP type is an error. This
557 * isn't specified by the RFC, but think about it..
558 */
559 if (*itp > NR_ICMP_TYPES ||
560 icmp_pointers[*itp].error)
561 goto out;
562 }
563 }
564
fdc0bde9
DL
565 sk = icmp_xmit_lock(net);
566 if (sk == NULL)
1da177e4
LT
567 return;
568
569 /*
570 * Construct source address and options.
571 */
572
573 saddr = iph->daddr;
1c2fb7f9 574 if (!(rt->rt_flags & RTCF_LOCAL)) {
6e1d9103
PM
575 struct net_device *dev = NULL;
576
685c7944 577 rcu_read_lock();
c7537967
DM
578 if (rt_is_input_route(rt) &&
579 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr)
92101b3b 580 dev = dev_get_by_index_rcu(net, inet_iif(skb_in));
6e1d9103 581
685c7944 582 if (dev)
6e1d9103 583 saddr = inet_select_addr(dev, 0, RT_SCOPE_LINK);
685c7944 584 else
1c2fb7f9 585 saddr = 0;
685c7944 586 rcu_read_unlock();
1c2fb7f9 587 }
1da177e4
LT
588
589 tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) |
590 IPTOS_PREC_INTERNETCONTROL) :
591 iph->tos;
6fa3eb70 592 mark = IP4_REPLY_MARK(net, skb_in->mark);
1da177e4 593
f6d8bd05 594 if (ip_options_echo(&icmp_param.replyopts.opt.opt, skb_in))
fa60cf7f 595 goto out_unlock;
1da177e4
LT
596
597
598 /*
599 * Prepare data for ICMP header.
600 */
601
602 icmp_param.data.icmph.type = type;
603 icmp_param.data.icmph.code = code;
604 icmp_param.data.icmph.un.gateway = info;
605 icmp_param.data.icmph.checksum = 0;
606 icmp_param.skb = skb_in;
d56f90a7 607 icmp_param.offset = skb_network_offset(skb_in);
405666db 608 inet_sk(sk)->tos = tos;
6fa3eb70 609 sk->sk_mark = mark;
1da177e4 610 ipc.addr = iph->saddr;
f6d8bd05 611 ipc.opt = &icmp_param.replyopts.opt;
2244d07b 612 ipc.tx_flags = 0;
1da177e4 613
6fa3eb70 614 rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, mark,
f6d460cf
DM
615 type, code, &icmp_param);
616 if (IS_ERR(rt))
617 goto out_unlock;
1da177e4 618
a48eff12 619 if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
1da177e4
LT
620 goto ende;
621
622 /* RFC says return as much as we can without exceeding 576 bytes. */
623
d8d1f30b 624 room = dst_mtu(&rt->dst);
1da177e4
LT
625 if (room > 576)
626 room = 576;
f6d8bd05 627 room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
1da177e4
LT
628 room -= sizeof(struct icmphdr);
629
630 icmp_param.data_len = skb_in->len - icmp_param.offset;
631 if (icmp_param.data_len > room)
632 icmp_param.data_len = room;
633 icmp_param.head_len = sizeof(struct icmphdr);
634
77968b78 635 icmp_push_reply(&icmp_param, &fl4, &ipc, &rt);
1da177e4
LT
636ende:
637 ip_rt_put(rt);
638out_unlock:
405666db 639 icmp_xmit_unlock(sk);
1da177e4
LT
640out:;
641}
4bc2f18b 642EXPORT_SYMBOL(icmp_send);
1da177e4
LT
643
644
1de9243b
DM
645static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
646{
647 const struct iphdr *iph = (const struct iphdr *) skb->data;
648 const struct net_protocol *ipprot;
649 int protocol = iph->protocol;
650
f0a70e90
DM
651 /* Checkin full IP header plus 8 bytes of protocol to
652 * avoid additional coding at protocol handlers.
653 */
654 if (!pskb_may_pull(skb, iph->ihl * 4 + 8))
655 return;
656
1de9243b
DM
657 raw_icmp_error(skb, protocol, info);
658
659 rcu_read_lock();
660 ipprot = rcu_dereference(inet_protos[protocol]);
661 if (ipprot && ipprot->err_handler)
662 ipprot->err_handler(skb, info);
663 rcu_read_unlock();
664}
665
1da177e4
LT
666/*
667 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
668 */
669
670static void icmp_unreach(struct sk_buff *skb)
671{
b71d1d42 672 const struct iphdr *iph;
1da177e4 673 struct icmphdr *icmph;
dde1bc0e 674 struct net *net;
f9242b6b 675 u32 info = 0;
dde1bc0e 676
adf30907 677 net = dev_net(skb_dst(skb)->dev);
1da177e4
LT
678
679 /*
680 * Incomplete header ?
681 * Only checks for the IP header, there should be an
682 * additional check for longer headers in upper levels.
683 */
684
685 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
686 goto out_err;
687
88c7664f 688 icmph = icmp_hdr(skb);
b71d1d42 689 iph = (const struct iphdr *)skb->data;
1da177e4
LT
690
691 if (iph->ihl < 5) /* Mangled header, drop. */
692 goto out_err;
693
694 if (icmph->type == ICMP_DEST_UNREACH) {
695 switch (icmph->code & 15) {
696 case ICMP_NET_UNREACH:
697 case ICMP_HOST_UNREACH:
698 case ICMP_PROT_UNREACH:
699 case ICMP_PORT_UNREACH:
700 break;
701 case ICMP_FRAG_NEEDED:
702 if (ipv4_config.no_pmtu_disc) {
afd46503 703 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: fragmentation needed and DF set\n"),
673d57e7 704 &iph->daddr);
1da177e4 705 } else {
46517008 706 info = ntohs(icmph->un.frag.mtu);
1da177e4
LT
707 }
708 break;
709 case ICMP_SR_FAILED:
afd46503 710 LIMIT_NETDEBUG(KERN_INFO pr_fmt("%pI4: Source Route Failed\n"),
673d57e7 711 &iph->daddr);
1da177e4
LT
712 break;
713 default:
714 break;
715 }
716 if (icmph->code > NR_ICMP_UNREACH)
717 goto out;
718 } else if (icmph->type == ICMP_PARAMETERPROB)
719 info = ntohl(icmph->un.gateway) >> 24;
720
721 /*
722 * Throw it at our lower layers
723 *
724 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
725 * header.
726 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
727 * transport layer.
728 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
729 * transport layer.
730 */
731
732 /*
25985edc 733 * Check the other end isn't violating RFC 1122. Some routers send
1da177e4
LT
734 * bogus responses to broadcast frames. If you see this message
735 * first check your netmask matches at both ends, if it does then
736 * get the other vendor to fix their kit.
737 */
738
b34a95ee 739 if (!net->ipv4.sysctl_icmp_ignore_bogus_error_responses &&
dde1bc0e 740 inet_addr_type(net, iph->daddr) == RTN_BROADCAST) {
e87cc472
JP
741 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
742 &ip_hdr(skb)->saddr,
743 icmph->type, icmph->code,
744 &iph->daddr, skb->dev->name);
1da177e4
LT
745 goto out;
746 }
747
1de9243b 748 icmp_socket_deliver(skb, info);
1da177e4
LT
749
750out:
751 return;
752out_err:
dcfc23ca 753 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
1da177e4
LT
754 goto out;
755}
756
757
758/*
759 * Handle ICMP_REDIRECT.
760 */
761
762static void icmp_redirect(struct sk_buff *skb)
763{
94206125
DM
764 if (skb->len < sizeof(struct iphdr)) {
765 ICMP_INC_STATS_BH(dev_net(skb->dev), ICMP_MIB_INERRORS);
766 return;
767 }
1da177e4 768
1da177e4 769 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
94206125 770 return;
c319b4d7 771
d3351b75 772 icmp_socket_deliver(skb, icmp_hdr(skb)->un.gateway);
1da177e4
LT
773}
774
775/*
776 * Handle ICMP_ECHO ("ping") requests.
777 *
778 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
779 * requests.
780 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
781 * included in the reply.
782 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
783 * echo requests, MUST have default=NOT.
784 * See also WRT handling of options once they are done and working.
785 */
786
787static void icmp_echo(struct sk_buff *skb)
788{
b34a95ee
PE
789 struct net *net;
790
adf30907 791 net = dev_net(skb_dst(skb)->dev);
b34a95ee 792 if (!net->ipv4.sysctl_icmp_echo_ignore_all) {
1da177e4
LT
793 struct icmp_bxm icmp_param;
794
88c7664f 795 icmp_param.data.icmph = *icmp_hdr(skb);
1da177e4
LT
796 icmp_param.data.icmph.type = ICMP_ECHOREPLY;
797 icmp_param.skb = skb;
798 icmp_param.offset = 0;
799 icmp_param.data_len = skb->len;
800 icmp_param.head_len = sizeof(struct icmphdr);
801 icmp_reply(&icmp_param, skb);
802 }
803}
804
805/*
806 * Handle ICMP Timestamp requests.
807 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
808 * SHOULD be in the kernel for minimum random latency.
809 * MUST be accurate to a few minutes.
810 * MUST be updated at least at 15Hz.
811 */
812static void icmp_timestamp(struct sk_buff *skb)
813{
f25c3d61 814 struct timespec tv;
1da177e4
LT
815 struct icmp_bxm icmp_param;
816 /*
817 * Too short.
818 */
819 if (skb->len < 4)
820 goto out_err;
821
822 /*
823 * Fill in the current time as ms since midnight UT:
824 */
f25c3d61
YH
825 getnstimeofday(&tv);
826 icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC +
827 tv.tv_nsec / NSEC_PER_MSEC);
1da177e4
LT
828 icmp_param.data.times[2] = icmp_param.data.times[1];
829 if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
830 BUG();
88c7664f 831 icmp_param.data.icmph = *icmp_hdr(skb);
1da177e4
LT
832 icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
833 icmp_param.data.icmph.code = 0;
834 icmp_param.skb = skb;
835 icmp_param.offset = 0;
836 icmp_param.data_len = 0;
837 icmp_param.head_len = sizeof(struct icmphdr) + 12;
838 icmp_reply(&icmp_param, skb);
839out:
840 return;
841out_err:
adf30907 842 ICMP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
1da177e4
LT
843 goto out;
844}
845
1da177e4
LT
846static void icmp_discard(struct sk_buff *skb)
847{
848}
849
850/*
851 * Deal with incoming ICMP packets.
852 */
853int icmp_rcv(struct sk_buff *skb)
854{
855 struct icmphdr *icmph;
511c3f92 856 struct rtable *rt = skb_rtable(skb);
d8d1f30b 857 struct net *net = dev_net(rt->dst.dev);
6fa3eb70
S
858 /*mtk_net: save ping reply sk*/
859 struct sock *ping_sk = NULL;
860
aebcf82c 861 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
def8b4fa 862 struct sec_path *sp = skb_sec_path(skb);
8b7817f3
HX
863 int nh;
864
def8b4fa 865 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
aebcf82c
HX
866 XFRM_STATE_ICMP))
867 goto drop;
868
8b7817f3
HX
869 if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
870 goto drop;
871
872 nh = skb_network_offset(skb);
873 skb_set_network_header(skb, sizeof(*icmph));
874
875 if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
876 goto drop;
877
878 skb_set_network_header(skb, nh);
879 }
880
dcfc23ca 881 ICMP_INC_STATS_BH(net, ICMP_MIB_INMSGS);
1da177e4
LT
882
883 switch (skb->ip_summed) {
84fa7933 884 case CHECKSUM_COMPLETE:
d3bc23e7 885 if (!csum_fold(skb->csum))
1da177e4 886 break;
fb286bb2 887 /* fall through */
1da177e4 888 case CHECKSUM_NONE:
fb286bb2
HX
889 skb->csum = 0;
890 if (__skb_checksum_complete(skb))
6a5dc9e5 891 goto csum_error;
1da177e4
LT
892 }
893
8cf22943
HX
894 if (!pskb_pull(skb, sizeof(*icmph)))
895 goto error;
1da177e4 896
88c7664f 897 icmph = icmp_hdr(skb);
1da177e4 898
f66ac03d 899 ICMPMSGIN_INC_STATS_BH(net, icmph->type);
1da177e4
LT
900 /*
901 * 18 is the highest 'known' ICMP type. Anything else is a mystery
902 *
903 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
904 * discarded.
905 */
906 if (icmph->type > NR_ICMP_TYPES)
907 goto error;
908
909
910 /*
911 * Parse the ICMP message
912 */
913
e905a9ed 914 if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
1da177e4
LT
915 /*
916 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
917 * silently ignored (we let user decide with a sysctl).
918 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
919 * discarded if to broadcast/multicast.
920 */
4c866aa7
AK
921 if ((icmph->type == ICMP_ECHO ||
922 icmph->type == ICMP_TIMESTAMP) &&
b34a95ee 923 net->ipv4.sysctl_icmp_echo_ignore_broadcasts) {
1da177e4
LT
924 goto error;
925 }
926 if (icmph->type != ICMP_ECHO &&
927 icmph->type != ICMP_TIMESTAMP &&
928 icmph->type != ICMP_ADDRESS &&
929 icmph->type != ICMP_ADDRESSREPLY) {
930 goto error;
e905a9ed 931 }
1da177e4
LT
932 }
933
1da177e4 934 icmp_pointers[icmph->type].handler(skb);
6fa3eb70
S
935
936 if(icmph->type == ICMP_ECHOREPLY && skb->sk != NULL ){
937 ping_sk = skb->sk;
938 }
939
1da177e4 940drop:
6fa3eb70
S
941 if(ping_sk){
942 kfree_skb(skb);
943 sock_put(ping_sk);
944 } else {
945 kfree_skb(skb);
946 }
1da177e4 947 return 0;
6a5dc9e5
ED
948csum_error:
949 ICMP_INC_STATS_BH(net, ICMP_MIB_CSUMERRORS);
1da177e4 950error:
dcfc23ca 951 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
1da177e4
LT
952 goto drop;
953}
954
5b052042
LW
955void icmp_err(struct sk_buff *skb, u32 info)
956{
957 struct iphdr *iph = (struct iphdr *)skb->data;
6fa3eb70
S
958 int offset = iph->ihl<<2;
959 struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset);
5b052042
LW
960 int type = icmp_hdr(skb)->type;
961 int code = icmp_hdr(skb)->code;
962 struct net *net = dev_net(skb->dev);
963
964 /*
965 * Use ping_err to handle all icmp errors except those
966 * triggered by ICMP_ECHOREPLY which sent from kernel.
967 */
968 if (icmph->type != ICMP_ECHOREPLY) {
6fa3eb70 969 ping_err(skb, offset, info);
5b052042
LW
970 return;
971 }
972
973 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
974 ipv4_update_pmtu(skb, net, info, 0, 0, IPPROTO_ICMP, 0);
975 else if (type == ICMP_REDIRECT)
976 ipv4_redirect(skb, net, 0, 0, IPPROTO_ICMP, 0);
977}
978
1da177e4
LT
979/*
980 * This table is the definition of how we handle ICMP.
981 */
9b5b5cff 982static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
1da177e4 983 [ICMP_ECHOREPLY] = {
c319b4d7 984 .handler = ping_rcv,
1da177e4
LT
985 },
986 [1] = {
1da177e4
LT
987 .handler = icmp_discard,
988 .error = 1,
989 },
990 [2] = {
1da177e4
LT
991 .handler = icmp_discard,
992 .error = 1,
993 },
994 [ICMP_DEST_UNREACH] = {
1da177e4
LT
995 .handler = icmp_unreach,
996 .error = 1,
997 },
998 [ICMP_SOURCE_QUENCH] = {
1da177e4
LT
999 .handler = icmp_unreach,
1000 .error = 1,
1001 },
1002 [ICMP_REDIRECT] = {
1da177e4
LT
1003 .handler = icmp_redirect,
1004 .error = 1,
1005 },
1006 [6] = {
1da177e4
LT
1007 .handler = icmp_discard,
1008 .error = 1,
1009 },
1010 [7] = {
1da177e4
LT
1011 .handler = icmp_discard,
1012 .error = 1,
1013 },
1014 [ICMP_ECHO] = {
1da177e4
LT
1015 .handler = icmp_echo,
1016 },
1017 [9] = {
1da177e4
LT
1018 .handler = icmp_discard,
1019 .error = 1,
1020 },
1021 [10] = {
1da177e4
LT
1022 .handler = icmp_discard,
1023 .error = 1,
1024 },
1025 [ICMP_TIME_EXCEEDED] = {
1da177e4
LT
1026 .handler = icmp_unreach,
1027 .error = 1,
1028 },
1029 [ICMP_PARAMETERPROB] = {
1da177e4
LT
1030 .handler = icmp_unreach,
1031 .error = 1,
1032 },
1033 [ICMP_TIMESTAMP] = {
1da177e4
LT
1034 .handler = icmp_timestamp,
1035 },
1036 [ICMP_TIMESTAMPREPLY] = {
1da177e4
LT
1037 .handler = icmp_discard,
1038 },
1039 [ICMP_INFO_REQUEST] = {
1da177e4
LT
1040 .handler = icmp_discard,
1041 },
e905a9ed 1042 [ICMP_INFO_REPLY] = {
1da177e4
LT
1043 .handler = icmp_discard,
1044 },
1045 [ICMP_ADDRESS] = {
838942a5 1046 .handler = icmp_discard,
1da177e4
LT
1047 },
1048 [ICMP_ADDRESSREPLY] = {
838942a5 1049 .handler = icmp_discard,
1da177e4
LT
1050 },
1051};
1052
4a6ad7a1 1053static void __net_exit icmp_sk_exit(struct net *net)
1da177e4 1054{
1da177e4
LT
1055 int i;
1056
5c8cafd6 1057 for_each_possible_cpu(i)
c1e9894d 1058 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
4a6ad7a1
DL
1059 kfree(net->ipv4.icmp_sk);
1060 net->ipv4.icmp_sk = NULL;
a5710d65 1061}
1da177e4 1062
263173af 1063static int __net_init icmp_sk_init(struct net *net)
a5710d65 1064{
a5710d65
DL
1065 int i, err;
1066
4a6ad7a1
DL
1067 net->ipv4.icmp_sk =
1068 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
1069 if (net->ipv4.icmp_sk == NULL)
79c91159
DL
1070 return -ENOMEM;
1071
a5710d65 1072 for_each_possible_cpu(i) {
1e3cf683 1073 struct sock *sk;
1da177e4 1074
c1e9894d
DL
1075 err = inet_ctl_sock_create(&sk, PF_INET,
1076 SOCK_RAW, IPPROTO_ICMP, net);
1da177e4 1077 if (err < 0)
a5710d65 1078 goto fail;
1da177e4 1079
c1e9894d 1080 net->ipv4.icmp_sk[i] = sk;
1da177e4
LT
1081
1082 /* Enough space for 2 64K ICMP packets, including
87fb4b7b 1083 * sk_buff/skb_shared_info struct overhead.
1da177e4 1084 */
87fb4b7b 1085 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
1da177e4 1086
b3a5b6cc
ED
1087 /*
1088 * Speedup sock_wfree()
1089 */
1090 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
c1e9894d 1091 inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
1da177e4 1092 }
a24022e1
PE
1093
1094 /* Control parameters for ECHO replies. */
1095 net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1096 net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1097
1098 /* Control parameter - ignore bogus broadcast responses? */
1099 net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1100
1101 /*
1102 * Configurable global rate limit.
1103 *
1104 * ratelimit defines tokens/packet consumed for dst->rate_token
1105 * bucket ratemask defines which icmp types are ratelimited by
1106 * setting it's bit position.
1107 *
1108 * default:
1109 * dest unreachable (3), source quench (4),
1110 * time exceeded (11), parameter problem (12)
1111 */
1112
1113 net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1114 net->ipv4.sysctl_icmp_ratemask = 0x1818;
1115 net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1116
a5710d65
DL
1117 return 0;
1118
1119fail:
1d1c8d13 1120 for_each_possible_cpu(i)
c1e9894d 1121 inet_ctl_sock_destroy(net->ipv4.icmp_sk[i]);
1d1c8d13 1122 kfree(net->ipv4.icmp_sk);
a5710d65 1123 return err;
1da177e4
LT
1124}
1125
4a6ad7a1
DL
1126static struct pernet_operations __net_initdata icmp_sk_ops = {
1127 .init = icmp_sk_init,
1128 .exit = icmp_sk_exit,
1129};
1130
1131int __init icmp_init(void)
1132{
959d2726 1133 return register_pernet_subsys(&icmp_sk_ops);
4a6ad7a1 1134}