2 * NET3: Implementation of the ICMP protocol layer.
4 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
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.
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.
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
24 * Alan Cox : Added 216,128 byte paths to the MTU
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
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
42 * Thomas Quinot : ICMP Dest Unreach codes up to 15 are
44 * Andi Kleen : Check all packet lengths properly
45 * and moved all kfree_skb() up to
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
55 * Tristan Greaves : Added sysctl option to ignore bogus
56 * broadcast responses from broken routers.
60 * - Should use skb_pull() instead of all the manual checking.
61 * This would also greatly simply some upper layer error handlers. --AK
65 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
74 #include <linux/inet.h>
75 #include <linux/inetdevice.h>
76 #include <linux/netdevice.h>
77 #include <linux/string.h>
78 #include <linux/netfilter_ipv4.h>
79 #include <linux/slab.h>
82 #include <net/route.h>
83 #include <net/protocol.h>
89 #include <linux/skbuff.h>
91 #include <linux/errno.h>
92 #include <linux/timer.h>
93 #include <linux/init.h>
94 #include <asm/uaccess.h>
95 #include <net/checksum.h>
97 #include <net/inet_common.h>
98 #include <net/ip_fib.h>
101 * Build xmit assembly blocks
110 struct icmphdr icmph
;
114 struct ip_options_data replyopts
;
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'. */
120 const struct icmp_err icmp_err_convert
[] = {
122 .errno
= ENETUNREACH
, /* ICMP_NET_UNREACH */
126 .errno
= EHOSTUNREACH
, /* ICMP_HOST_UNREACH */
130 .errno
= ENOPROTOOPT
/* ICMP_PROT_UNREACH */,
134 .errno
= ECONNREFUSED
, /* ICMP_PORT_UNREACH */
138 .errno
= EMSGSIZE
, /* ICMP_FRAG_NEEDED */
142 .errno
= EOPNOTSUPP
, /* ICMP_SR_FAILED */
146 .errno
= ENETUNREACH
, /* ICMP_NET_UNKNOWN */
150 .errno
= EHOSTDOWN
, /* ICMP_HOST_UNKNOWN */
154 .errno
= ENONET
, /* ICMP_HOST_ISOLATED */
158 .errno
= ENETUNREACH
, /* ICMP_NET_ANO */
162 .errno
= EHOSTUNREACH
, /* ICMP_HOST_ANO */
166 .errno
= ENETUNREACH
, /* ICMP_NET_UNR_TOS */
170 .errno
= EHOSTUNREACH
, /* ICMP_HOST_UNR_TOS */
174 .errno
= EHOSTUNREACH
, /* ICMP_PKT_FILTERED */
178 .errno
= EHOSTUNREACH
, /* ICMP_PREC_VIOLATION */
182 .errno
= EHOSTUNREACH
, /* ICMP_PREC_CUTOFF */
186 EXPORT_SYMBOL(icmp_err_convert
);
189 * ICMP control array. This specifies what to do with each ICMP.
192 struct icmp_control
{
193 void (*handler
)(struct sk_buff
*skb
);
194 short error
; /* This ICMP is classed as an error message */
197 static const struct icmp_control icmp_pointers
[NR_ICMP_TYPES
+1];
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.
204 * On SMP we have one ICMP socket per-cpu.
206 static struct sock
*icmp_sk(struct net
*net
)
208 return net
->ipv4
.icmp_sk
[smp_processor_id()];
211 static inline struct sock
*icmp_xmit_lock(struct net
*net
)
219 if (unlikely(!spin_trylock(&sk
->sk_lock
.slock
))) {
220 /* This can happen if the output path signals a
221 * dst_link_failure() for an outgoing ICMP packet.
229 static inline void icmp_xmit_unlock(struct sock
*sk
)
231 spin_unlock_bh(&sk
->sk_lock
.slock
);
235 * Send an ICMP frame.
238 static inline bool icmpv4_xrlim_allow(struct net
*net
, struct rtable
*rt
,
239 struct flowi4
*fl4
, int type
, int code
)
241 struct dst_entry
*dst
= &rt
->dst
;
244 if (type
> NR_ICMP_TYPES
)
247 /* Don't limit PMTU discovery. */
248 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
)
251 /* No rate limit on loopback */
252 if (dst
->dev
&& (dst
->dev
->flags
&IFF_LOOPBACK
))
255 /* Limit if icmp type is enabled in ratemask. */
256 if ((1 << type
) & net
->ipv4
.sysctl_icmp_ratemask
) {
257 struct inet_peer
*peer
= inet_getpeer_v4(net
->ipv4
.peers
, fl4
->daddr
, 1);
258 rc
= inet_peer_xrlim_allow(peer
,
259 net
->ipv4
.sysctl_icmp_ratelimit
);
267 * Maintain the counters used in the SNMP statistics for outgoing ICMP
269 void icmp_out_count(struct net
*net
, unsigned char type
)
271 ICMPMSGOUT_INC_STATS(net
, type
);
272 ICMP_INC_STATS(net
, ICMP_MIB_OUTMSGS
);
276 * Checksum each fragment, and on the first include the headers and final
279 static int icmp_glue_bits(void *from
, char *to
, int offset
, int len
, int odd
,
282 struct icmp_bxm
*icmp_param
= (struct icmp_bxm
*)from
;
285 csum
= skb_copy_and_csum_bits(icmp_param
->skb
,
286 icmp_param
->offset
+ offset
,
289 skb
->csum
= csum_block_add(skb
->csum
, csum
, odd
);
290 if (icmp_pointers
[icmp_param
->data
.icmph
.type
].error
)
291 nf_ct_attach(skb
, icmp_param
->skb
);
295 static void icmp_push_reply(struct icmp_bxm
*icmp_param
,
297 struct ipcm_cookie
*ipc
, struct rtable
**rt
)
302 sk
= icmp_sk(dev_net((*rt
)->dst
.dev
));
303 if (ip_append_data(sk
, fl4
, icmp_glue_bits
, icmp_param
,
304 icmp_param
->data_len
+icmp_param
->head_len
,
305 icmp_param
->head_len
,
306 ipc
, rt
, MSG_DONTWAIT
) < 0) {
307 ICMP_INC_STATS_BH(sock_net(sk
), ICMP_MIB_OUTERRORS
);
308 ip_flush_pending_frames(sk
);
309 } else if ((skb
= skb_peek(&sk
->sk_write_queue
)) != NULL
) {
310 struct icmphdr
*icmph
= icmp_hdr(skb
);
312 struct sk_buff
*skb1
;
314 skb_queue_walk(&sk
->sk_write_queue
, skb1
) {
315 csum
= csum_add(csum
, skb1
->csum
);
317 csum
= csum_partial_copy_nocheck((void *)&icmp_param
->data
,
319 icmp_param
->head_len
, csum
);
320 icmph
->checksum
= csum_fold(csum
);
321 skb
->ip_summed
= CHECKSUM_NONE
;
322 ip_push_pending_frames(sk
, fl4
);
327 * Driving logic for building and sending ICMP messages.
330 static void icmp_reply(struct icmp_bxm
*icmp_param
, struct sk_buff
*skb
)
332 struct ipcm_cookie ipc
;
333 struct rtable
*rt
= skb_rtable(skb
);
334 struct net
*net
= dev_net(rt
->dst
.dev
);
337 struct inet_sock
*inet
;
340 if (ip_options_echo(&icmp_param
->replyopts
.opt
.opt
, skb
))
343 sk
= icmp_xmit_lock(net
);
348 icmp_param
->data
.icmph
.checksum
= 0;
350 inet
->tos
= ip_hdr(skb
)->tos
;
351 daddr
= ipc
.addr
= ip_hdr(skb
)->saddr
;
352 saddr
= fib_compute_spec_dst(skb
);
355 if (icmp_param
->replyopts
.opt
.opt
.optlen
) {
356 ipc
.opt
= &icmp_param
->replyopts
.opt
;
357 if (ipc
.opt
->opt
.srr
)
358 daddr
= icmp_param
->replyopts
.opt
.opt
.faddr
;
360 memset(&fl4
, 0, sizeof(fl4
));
363 fl4
.flowi4_tos
= RT_TOS(ip_hdr(skb
)->tos
);
364 fl4
.flowi4_proto
= IPPROTO_ICMP
;
365 security_skb_classify_flow(skb
, flowi4_to_flowi(&fl4
));
366 rt
= ip_route_output_key(net
, &fl4
);
369 if (icmpv4_xrlim_allow(net
, rt
, &fl4
, icmp_param
->data
.icmph
.type
,
370 icmp_param
->data
.icmph
.code
))
371 icmp_push_reply(icmp_param
, &fl4
, &ipc
, &rt
);
374 icmp_xmit_unlock(sk
);
377 static struct rtable
*icmp_route_lookup(struct net
*net
,
379 struct sk_buff
*skb_in
,
380 const struct iphdr
*iph
,
381 __be32 saddr
, u8 tos
,
383 struct icmp_bxm
*param
)
385 struct rtable
*rt
, *rt2
;
386 struct flowi4 fl4_dec
;
389 memset(fl4
, 0, sizeof(*fl4
));
390 fl4
->daddr
= (param
->replyopts
.opt
.opt
.srr
?
391 param
->replyopts
.opt
.opt
.faddr
: iph
->saddr
);
393 fl4
->flowi4_tos
= RT_TOS(tos
);
394 fl4
->flowi4_proto
= IPPROTO_ICMP
;
395 fl4
->fl4_icmp_type
= type
;
396 fl4
->fl4_icmp_code
= code
;
397 security_skb_classify_flow(skb_in
, flowi4_to_flowi(fl4
));
398 rt
= __ip_route_output_key(net
, fl4
);
402 /* No need to clone since we're just using its address. */
405 rt
= (struct rtable
*) xfrm_lookup(net
, &rt
->dst
,
406 flowi4_to_flowi(fl4
), NULL
, 0);
410 } else if (PTR_ERR(rt
) == -EPERM
) {
415 err
= xfrm_decode_session_reverse(skb_in
, flowi4_to_flowi(&fl4_dec
), AF_INET
);
417 goto relookup_failed
;
419 if (inet_addr_type(net
, fl4_dec
.saddr
) == RTN_LOCAL
) {
420 rt2
= __ip_route_output_key(net
, &fl4_dec
);
424 struct flowi4 fl4_2
= {};
425 unsigned long orefdst
;
427 fl4_2
.daddr
= fl4_dec
.saddr
;
428 rt2
= ip_route_output_key(net
, &fl4_2
);
431 goto relookup_failed
;
434 orefdst
= skb_in
->_skb_refdst
; /* save old refdst */
435 err
= ip_route_input(skb_in
, fl4_dec
.daddr
, fl4_dec
.saddr
,
436 RT_TOS(tos
), rt2
->dst
.dev
);
438 dst_release(&rt2
->dst
);
439 rt2
= skb_rtable(skb_in
);
440 skb_in
->_skb_refdst
= orefdst
; /* restore old refdst */
444 goto relookup_failed
;
446 rt2
= (struct rtable
*) xfrm_lookup(net
, &rt2
->dst
,
447 flowi4_to_flowi(&fl4_dec
), NULL
,
450 dst_release(&rt
->dst
);
451 memcpy(fl4
, &fl4_dec
, sizeof(*fl4
));
453 } else if (PTR_ERR(rt2
) == -EPERM
) {
455 dst_release(&rt
->dst
);
459 goto relookup_failed
;
470 * Send an ICMP message in response to a situation
472 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
473 * MAY send more (we do).
474 * MUST NOT change this header information.
475 * MUST NOT reply to a multicast/broadcast IP address.
476 * MUST NOT reply to a multicast/broadcast MAC address.
477 * MUST reply to only the first fragment.
480 void icmp_send(struct sk_buff
*skb_in
, int type
, int code
, __be32 info
)
484 struct icmp_bxm icmp_param
;
485 struct rtable
*rt
= skb_rtable(skb_in
);
486 struct ipcm_cookie ipc
;
495 net
= dev_net(rt
->dst
.dev
);
498 * Find the original header. It is expected to be valid, of course.
499 * Check this, icmp_send is called from the most obscure devices
502 iph
= ip_hdr(skb_in
);
504 if ((u8
*)iph
< skb_in
->head
||
505 (skb_in
->network_header
+ sizeof(*iph
)) > skb_in
->tail
)
509 * No replies to physical multicast/broadcast
511 if (skb_in
->pkt_type
!= PACKET_HOST
)
515 * Now check at the protocol level
517 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))
521 * Only reply to fragment 0. We byte re-order the constant
522 * mask for efficiency.
524 if (iph
->frag_off
& htons(IP_OFFSET
))
528 * If we send an ICMP error to an ICMP error a mess would result..
530 if (icmp_pointers
[type
].error
) {
532 * We are an error, check if we are replying to an
535 if (iph
->protocol
== IPPROTO_ICMP
) {
536 u8 _inner_type
, *itp
;
538 itp
= skb_header_pointer(skb_in
,
539 skb_network_header(skb_in
) +
541 offsetof(struct icmphdr
,
550 * Assume any unknown ICMP type is an error. This
551 * isn't specified by the RFC, but think about it..
553 if (*itp
> NR_ICMP_TYPES
||
554 icmp_pointers
[*itp
].error
)
559 sk
= icmp_xmit_lock(net
);
564 * Construct source address and options.
568 if (!(rt
->rt_flags
& RTCF_LOCAL
)) {
569 struct net_device
*dev
= NULL
;
572 if (rt_is_input_route(rt
) &&
573 net
->ipv4
.sysctl_icmp_errors_use_inbound_ifaddr
)
574 dev
= dev_get_by_index_rcu(net
, inet_iif(skb_in
));
577 saddr
= inet_select_addr(dev
, 0, RT_SCOPE_LINK
);
583 tos
= icmp_pointers
[type
].error
? ((iph
->tos
& IPTOS_TOS_MASK
) |
584 IPTOS_PREC_INTERNETCONTROL
) :
587 if (ip_options_echo(&icmp_param
.replyopts
.opt
.opt
, skb_in
))
592 * Prepare data for ICMP header.
595 icmp_param
.data
.icmph
.type
= type
;
596 icmp_param
.data
.icmph
.code
= code
;
597 icmp_param
.data
.icmph
.un
.gateway
= info
;
598 icmp_param
.data
.icmph
.checksum
= 0;
599 icmp_param
.skb
= skb_in
;
600 icmp_param
.offset
= skb_network_offset(skb_in
);
601 inet_sk(sk
)->tos
= tos
;
602 ipc
.addr
= iph
->saddr
;
603 ipc
.opt
= &icmp_param
.replyopts
.opt
;
606 rt
= icmp_route_lookup(net
, &fl4
, skb_in
, iph
, saddr
, tos
,
607 type
, code
, &icmp_param
);
611 if (!icmpv4_xrlim_allow(net
, rt
, &fl4
, type
, code
))
614 /* RFC says return as much as we can without exceeding 576 bytes. */
616 room
= dst_mtu(&rt
->dst
);
619 room
-= sizeof(struct iphdr
) + icmp_param
.replyopts
.opt
.opt
.optlen
;
620 room
-= sizeof(struct icmphdr
);
622 icmp_param
.data_len
= skb_in
->len
- icmp_param
.offset
;
623 if (icmp_param
.data_len
> room
)
624 icmp_param
.data_len
= room
;
625 icmp_param
.head_len
= sizeof(struct icmphdr
);
627 icmp_push_reply(&icmp_param
, &fl4
, &ipc
, &rt
);
631 icmp_xmit_unlock(sk
);
634 EXPORT_SYMBOL(icmp_send
);
637 static void icmp_socket_deliver(struct sk_buff
*skb
, u32 info
)
639 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
640 const struct net_protocol
*ipprot
;
641 int protocol
= iph
->protocol
;
643 /* Checkin full IP header plus 8 bytes of protocol to
644 * avoid additional coding at protocol handlers.
646 if (!pskb_may_pull(skb
, iph
->ihl
* 4 + 8))
649 raw_icmp_error(skb
, protocol
, info
);
652 ipprot
= rcu_dereference(inet_protos
[protocol
]);
653 if (ipprot
&& ipprot
->err_handler
)
654 ipprot
->err_handler(skb
, info
);
659 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
662 static void icmp_unreach(struct sk_buff
*skb
)
664 const struct iphdr
*iph
;
665 struct icmphdr
*icmph
;
669 net
= dev_net(skb_dst(skb
)->dev
);
672 * Incomplete header ?
673 * Only checks for the IP header, there should be an
674 * additional check for longer headers in upper levels.
677 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
680 icmph
= icmp_hdr(skb
);
681 iph
= (const struct iphdr
*)skb
->data
;
683 if (iph
->ihl
< 5) /* Mangled header, drop. */
686 if (icmph
->type
== ICMP_DEST_UNREACH
) {
687 switch (icmph
->code
& 15) {
688 case ICMP_NET_UNREACH
:
689 case ICMP_HOST_UNREACH
:
690 case ICMP_PROT_UNREACH
:
691 case ICMP_PORT_UNREACH
:
693 case ICMP_FRAG_NEEDED
:
694 if (ipv4_config
.no_pmtu_disc
) {
695 LIMIT_NETDEBUG(KERN_INFO
pr_fmt("%pI4: fragmentation needed and DF set\n"),
698 info
= ntohs(icmph
->un
.frag
.mtu
);
704 LIMIT_NETDEBUG(KERN_INFO
pr_fmt("%pI4: Source Route Failed\n"),
710 if (icmph
->code
> NR_ICMP_UNREACH
)
712 } else if (icmph
->type
== ICMP_PARAMETERPROB
)
713 info
= ntohl(icmph
->un
.gateway
) >> 24;
716 * Throw it at our lower layers
718 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
720 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
722 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
727 * Check the other end isn't violating RFC 1122. Some routers send
728 * bogus responses to broadcast frames. If you see this message
729 * first check your netmask matches at both ends, if it does then
730 * get the other vendor to fix their kit.
733 if (!net
->ipv4
.sysctl_icmp_ignore_bogus_error_responses
&&
734 inet_addr_type(net
, iph
->daddr
) == RTN_BROADCAST
) {
735 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
737 icmph
->type
, icmph
->code
,
738 &iph
->daddr
, skb
->dev
->name
);
742 icmp_socket_deliver(skb
, info
);
747 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
753 * Handle ICMP_REDIRECT.
756 static void icmp_redirect(struct sk_buff
*skb
)
758 if (skb
->len
< sizeof(struct iphdr
)) {
759 ICMP_INC_STATS_BH(dev_net(skb
->dev
), ICMP_MIB_INERRORS
);
763 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
766 icmp_socket_deliver(skb
, icmp_hdr(skb
)->un
.gateway
);
770 * Handle ICMP_ECHO ("ping") requests.
772 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
774 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
775 * included in the reply.
776 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
777 * echo requests, MUST have default=NOT.
778 * See also WRT handling of options once they are done and working.
781 static void icmp_echo(struct sk_buff
*skb
)
785 net
= dev_net(skb_dst(skb
)->dev
);
786 if (!net
->ipv4
.sysctl_icmp_echo_ignore_all
) {
787 struct icmp_bxm icmp_param
;
789 icmp_param
.data
.icmph
= *icmp_hdr(skb
);
790 icmp_param
.data
.icmph
.type
= ICMP_ECHOREPLY
;
791 icmp_param
.skb
= skb
;
792 icmp_param
.offset
= 0;
793 icmp_param
.data_len
= skb
->len
;
794 icmp_param
.head_len
= sizeof(struct icmphdr
);
795 icmp_reply(&icmp_param
, skb
);
800 * Handle ICMP Timestamp requests.
801 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
802 * SHOULD be in the kernel for minimum random latency.
803 * MUST be accurate to a few minutes.
804 * MUST be updated at least at 15Hz.
806 static void icmp_timestamp(struct sk_buff
*skb
)
809 struct icmp_bxm icmp_param
;
817 * Fill in the current time as ms since midnight UT:
820 icmp_param
.data
.times
[1] = htonl((tv
.tv_sec
% 86400) * MSEC_PER_SEC
+
821 tv
.tv_nsec
/ NSEC_PER_MSEC
);
822 icmp_param
.data
.times
[2] = icmp_param
.data
.times
[1];
823 if (skb_copy_bits(skb
, 0, &icmp_param
.data
.times
[0], 4))
825 icmp_param
.data
.icmph
= *icmp_hdr(skb
);
826 icmp_param
.data
.icmph
.type
= ICMP_TIMESTAMPREPLY
;
827 icmp_param
.data
.icmph
.code
= 0;
828 icmp_param
.skb
= skb
;
829 icmp_param
.offset
= 0;
830 icmp_param
.data_len
= 0;
831 icmp_param
.head_len
= sizeof(struct icmphdr
) + 12;
832 icmp_reply(&icmp_param
, skb
);
836 ICMP_INC_STATS_BH(dev_net(skb_dst(skb
)->dev
), ICMP_MIB_INERRORS
);
840 static void icmp_discard(struct sk_buff
*skb
)
845 * Deal with incoming ICMP packets.
847 int icmp_rcv(struct sk_buff
*skb
)
849 struct icmphdr
*icmph
;
850 struct rtable
*rt
= skb_rtable(skb
);
851 struct net
*net
= dev_net(rt
->dst
.dev
);
853 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
854 struct sec_path
*sp
= skb_sec_path(skb
);
857 if (!(sp
&& sp
->xvec
[sp
->len
- 1]->props
.flags
&
861 if (!pskb_may_pull(skb
, sizeof(*icmph
) + sizeof(struct iphdr
)))
864 nh
= skb_network_offset(skb
);
865 skb_set_network_header(skb
, sizeof(*icmph
));
867 if (!xfrm4_policy_check_reverse(NULL
, XFRM_POLICY_IN
, skb
))
870 skb_set_network_header(skb
, nh
);
873 ICMP_INC_STATS_BH(net
, ICMP_MIB_INMSGS
);
875 switch (skb
->ip_summed
) {
876 case CHECKSUM_COMPLETE
:
877 if (!csum_fold(skb
->csum
))
882 if (__skb_checksum_complete(skb
))
886 if (!pskb_pull(skb
, sizeof(*icmph
)))
889 icmph
= icmp_hdr(skb
);
891 ICMPMSGIN_INC_STATS_BH(net
, icmph
->type
);
893 * 18 is the highest 'known' ICMP type. Anything else is a mystery
895 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
898 if (icmph
->type
> NR_ICMP_TYPES
)
903 * Parse the ICMP message
906 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
908 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
909 * silently ignored (we let user decide with a sysctl).
910 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
911 * discarded if to broadcast/multicast.
913 if ((icmph
->type
== ICMP_ECHO
||
914 icmph
->type
== ICMP_TIMESTAMP
) &&
915 net
->ipv4
.sysctl_icmp_echo_ignore_broadcasts
) {
918 if (icmph
->type
!= ICMP_ECHO
&&
919 icmph
->type
!= ICMP_TIMESTAMP
&&
920 icmph
->type
!= ICMP_ADDRESS
&&
921 icmph
->type
!= ICMP_ADDRESSREPLY
) {
926 icmp_pointers
[icmph
->type
].handler(skb
);
932 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
937 * This table is the definition of how we handle ICMP.
939 static const struct icmp_control icmp_pointers
[NR_ICMP_TYPES
+ 1] = {
944 .handler
= icmp_discard
,
948 .handler
= icmp_discard
,
951 [ICMP_DEST_UNREACH
] = {
952 .handler
= icmp_unreach
,
955 [ICMP_SOURCE_QUENCH
] = {
956 .handler
= icmp_unreach
,
960 .handler
= icmp_redirect
,
964 .handler
= icmp_discard
,
968 .handler
= icmp_discard
,
972 .handler
= icmp_echo
,
975 .handler
= icmp_discard
,
979 .handler
= icmp_discard
,
982 [ICMP_TIME_EXCEEDED
] = {
983 .handler
= icmp_unreach
,
986 [ICMP_PARAMETERPROB
] = {
987 .handler
= icmp_unreach
,
991 .handler
= icmp_timestamp
,
993 [ICMP_TIMESTAMPREPLY
] = {
994 .handler
= icmp_discard
,
996 [ICMP_INFO_REQUEST
] = {
997 .handler
= icmp_discard
,
999 [ICMP_INFO_REPLY
] = {
1000 .handler
= icmp_discard
,
1003 .handler
= icmp_discard
,
1005 [ICMP_ADDRESSREPLY
] = {
1006 .handler
= icmp_discard
,
1010 static void __net_exit
icmp_sk_exit(struct net
*net
)
1014 for_each_possible_cpu(i
)
1015 inet_ctl_sock_destroy(net
->ipv4
.icmp_sk
[i
]);
1016 kfree(net
->ipv4
.icmp_sk
);
1017 net
->ipv4
.icmp_sk
= NULL
;
1020 static int __net_init
icmp_sk_init(struct net
*net
)
1025 kzalloc(nr_cpu_ids
* sizeof(struct sock
*), GFP_KERNEL
);
1026 if (net
->ipv4
.icmp_sk
== NULL
)
1029 for_each_possible_cpu(i
) {
1032 err
= inet_ctl_sock_create(&sk
, PF_INET
,
1033 SOCK_RAW
, IPPROTO_ICMP
, net
);
1037 net
->ipv4
.icmp_sk
[i
] = sk
;
1039 /* Enough space for 2 64K ICMP packets, including
1040 * sk_buff/skb_shared_info struct overhead.
1042 sk
->sk_sndbuf
= 2 * SKB_TRUESIZE(64 * 1024);
1045 * Speedup sock_wfree()
1047 sock_set_flag(sk
, SOCK_USE_WRITE_QUEUE
);
1048 inet_sk(sk
)->pmtudisc
= IP_PMTUDISC_DONT
;
1051 /* Control parameters for ECHO replies. */
1052 net
->ipv4
.sysctl_icmp_echo_ignore_all
= 0;
1053 net
->ipv4
.sysctl_icmp_echo_ignore_broadcasts
= 1;
1055 /* Control parameter - ignore bogus broadcast responses? */
1056 net
->ipv4
.sysctl_icmp_ignore_bogus_error_responses
= 1;
1059 * Configurable global rate limit.
1061 * ratelimit defines tokens/packet consumed for dst->rate_token
1062 * bucket ratemask defines which icmp types are ratelimited by
1063 * setting it's bit position.
1066 * dest unreachable (3), source quench (4),
1067 * time exceeded (11), parameter problem (12)
1070 net
->ipv4
.sysctl_icmp_ratelimit
= 1 * HZ
;
1071 net
->ipv4
.sysctl_icmp_ratemask
= 0x1818;
1072 net
->ipv4
.sysctl_icmp_errors_use_inbound_ifaddr
= 0;
1077 for_each_possible_cpu(i
)
1078 inet_ctl_sock_destroy(net
->ipv4
.icmp_sk
[i
]);
1079 kfree(net
->ipv4
.icmp_sk
);
1083 static struct pernet_operations __net_initdata icmp_sk_ops
= {
1084 .init
= icmp_sk_init
,
1085 .exit
= icmp_sk_exit
,
1088 int __init
icmp_init(void)
1090 return register_pernet_subsys(&icmp_sk_ops
);