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
);
268 * Maintain the counters used in the SNMP statistics for outgoing ICMP
270 void icmp_out_count(struct net
*net
, unsigned char type
)
272 ICMPMSGOUT_INC_STATS(net
, type
);
273 ICMP_INC_STATS(net
, ICMP_MIB_OUTMSGS
);
277 * Checksum each fragment, and on the first include the headers and final
280 static int icmp_glue_bits(void *from
, char *to
, int offset
, int len
, int odd
,
283 struct icmp_bxm
*icmp_param
= (struct icmp_bxm
*)from
;
286 csum
= skb_copy_and_csum_bits(icmp_param
->skb
,
287 icmp_param
->offset
+ offset
,
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
);
296 static void icmp_push_reply(struct icmp_bxm
*icmp_param
,
298 struct ipcm_cookie
*ipc
, struct rtable
**rt
)
303 sk
= icmp_sk(dev_net((*rt
)->dst
.dev
));
304 if (ip_append_data(sk
, fl4
, icmp_glue_bits
, icmp_param
,
305 icmp_param
->data_len
+icmp_param
->head_len
,
306 icmp_param
->head_len
,
307 ipc
, rt
, MSG_DONTWAIT
) < 0) {
308 ICMP_INC_STATS_BH(sock_net(sk
), ICMP_MIB_OUTERRORS
);
309 ip_flush_pending_frames(sk
);
310 } else if ((skb
= skb_peek(&sk
->sk_write_queue
)) != NULL
) {
311 struct icmphdr
*icmph
= icmp_hdr(skb
);
313 struct sk_buff
*skb1
;
315 skb_queue_walk(&sk
->sk_write_queue
, skb1
) {
316 csum
= csum_add(csum
, skb1
->csum
);
318 csum
= csum_partial_copy_nocheck((void *)&icmp_param
->data
,
320 icmp_param
->head_len
, csum
);
321 icmph
->checksum
= csum_fold(csum
);
322 skb
->ip_summed
= CHECKSUM_NONE
;
323 ip_push_pending_frames(sk
, fl4
);
328 * Driving logic for building and sending ICMP messages.
331 static void icmp_reply(struct icmp_bxm
*icmp_param
, struct sk_buff
*skb
)
333 struct ipcm_cookie ipc
;
334 struct rtable
*rt
= skb_rtable(skb
);
335 struct net
*net
= dev_net(rt
->dst
.dev
);
338 struct inet_sock
*inet
;
341 if (ip_options_echo(&icmp_param
->replyopts
.opt
.opt
, skb
))
344 sk
= icmp_xmit_lock(net
);
349 icmp_param
->data
.icmph
.checksum
= 0;
351 inet
->tos
= ip_hdr(skb
)->tos
;
352 daddr
= ipc
.addr
= ip_hdr(skb
)->saddr
;
353 saddr
= fib_compute_spec_dst(skb
);
359 if (icmp_param
->replyopts
.opt
.opt
.optlen
) {
360 ipc
.opt
= &icmp_param
->replyopts
.opt
;
361 if (ipc
.opt
->opt
.srr
)
362 daddr
= icmp_param
->replyopts
.opt
.opt
.faddr
;
364 memset(&fl4
, 0, sizeof(fl4
));
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
);
373 if (icmpv4_xrlim_allow(net
, rt
, &fl4
, icmp_param
->data
.icmph
.type
,
374 icmp_param
->data
.icmph
.code
))
375 icmp_push_reply(icmp_param
, &fl4
, &ipc
, &rt
);
378 icmp_xmit_unlock(sk
);
381 static struct rtable
*icmp_route_lookup(struct net
*net
,
383 struct sk_buff
*skb_in
,
384 const struct iphdr
*iph
,
385 __be32 saddr
, u8 tos
,
387 struct icmp_bxm
*param
)
389 struct rtable
*rt
, *rt2
;
390 struct flowi4 fl4_dec
;
393 memset(fl4
, 0, sizeof(*fl4
));
394 fl4
->daddr
= (param
->replyopts
.opt
.opt
.srr
?
395 param
->replyopts
.opt
.opt
.faddr
: iph
->saddr
);
397 fl4
->flowi4_tos
= RT_TOS(tos
);
398 fl4
->flowi4_proto
= IPPROTO_ICMP
;
399 fl4
->fl4_icmp_type
= type
;
400 fl4
->fl4_icmp_code
= code
;
401 security_skb_classify_flow(skb_in
, flowi4_to_flowi(fl4
));
402 rt
= __ip_route_output_key(net
, fl4
);
406 /* No need to clone since we're just using its address. */
409 rt
= (struct rtable
*) xfrm_lookup(net
, &rt
->dst
,
410 flowi4_to_flowi(fl4
), NULL
, 0);
414 } else if (PTR_ERR(rt
) == -EPERM
) {
419 err
= xfrm_decode_session_reverse(skb_in
, flowi4_to_flowi(&fl4_dec
), AF_INET
);
421 goto relookup_failed
;
423 if (inet_addr_type(net
, fl4_dec
.saddr
) == RTN_LOCAL
) {
424 rt2
= __ip_route_output_key(net
, &fl4_dec
);
428 struct flowi4 fl4_2
= {};
429 unsigned long orefdst
;
431 fl4_2
.daddr
= fl4_dec
.saddr
;
432 rt2
= ip_route_output_key(net
, &fl4_2
);
435 goto relookup_failed
;
438 orefdst
= skb_in
->_skb_refdst
; /* save old refdst */
439 err
= ip_route_input(skb_in
, fl4_dec
.daddr
, fl4_dec
.saddr
,
440 RT_TOS(tos
), rt2
->dst
.dev
);
442 dst_release(&rt2
->dst
);
443 rt2
= skb_rtable(skb_in
);
444 skb_in
->_skb_refdst
= orefdst
; /* restore old refdst */
448 goto relookup_failed
;
450 rt2
= (struct rtable
*) xfrm_lookup(net
, &rt2
->dst
,
451 flowi4_to_flowi(&fl4_dec
), NULL
,
454 dst_release(&rt
->dst
);
455 memcpy(fl4
, &fl4_dec
, sizeof(*fl4
));
457 } else if (PTR_ERR(rt2
) == -EPERM
) {
459 dst_release(&rt
->dst
);
463 goto relookup_failed
;
474 * Send an ICMP message in response to a situation
476 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
477 * MAY send more (we do).
478 * MUST NOT change this header information.
479 * MUST NOT reply to a multicast/broadcast IP address.
480 * MUST NOT reply to a multicast/broadcast MAC address.
481 * MUST reply to only the first fragment.
484 void icmp_send(struct sk_buff
*skb_in
, int type
, int code
, __be32 info
)
488 struct icmp_bxm
*icmp_param
;
489 struct rtable
*rt
= skb_rtable(skb_in
);
490 struct ipcm_cookie ipc
;
499 net
= dev_net(rt
->dst
.dev
);
502 * Find the original header. It is expected to be valid, of course.
503 * Check this, icmp_send is called from the most obscure devices
506 iph
= ip_hdr(skb_in
);
508 if ((u8
*)iph
< skb_in
->head
||
509 (skb_network_header(skb_in
) + sizeof(*iph
)) >
510 skb_tail_pointer(skb_in
))
514 * No replies to physical multicast/broadcast
516 if (skb_in
->pkt_type
!= PACKET_HOST
)
520 * Now check at the protocol level
522 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))
526 * Only reply to fragment 0. We byte re-order the constant
527 * mask for efficiency.
529 if (iph
->frag_off
& htons(IP_OFFSET
))
533 * If we send an ICMP error to an ICMP error a mess would result..
535 if (icmp_pointers
[type
].error
) {
537 * We are an error, check if we are replying to an
540 if (iph
->protocol
== IPPROTO_ICMP
) {
541 u8 _inner_type
, *itp
;
543 itp
= skb_header_pointer(skb_in
,
544 skb_network_header(skb_in
) +
546 offsetof(struct icmphdr
,
555 * Assume any unknown ICMP type is an error. This
556 * isn't specified by the RFC, but think about it..
558 if (*itp
> NR_ICMP_TYPES
||
559 icmp_pointers
[*itp
].error
)
564 icmp_param
= kmalloc(sizeof(*icmp_param
), GFP_ATOMIC
);
568 sk
= icmp_xmit_lock(net
);
573 * Construct source address and options.
577 if (!(rt
->rt_flags
& RTCF_LOCAL
)) {
578 struct net_device
*dev
= NULL
;
581 if (rt_is_input_route(rt
) &&
582 net
->ipv4
.sysctl_icmp_errors_use_inbound_ifaddr
)
583 dev
= dev_get_by_index_rcu(net
, inet_iif(skb_in
));
586 saddr
= inet_select_addr(dev
, 0, RT_SCOPE_LINK
);
592 tos
= icmp_pointers
[type
].error
? ((iph
->tos
& IPTOS_TOS_MASK
) |
593 IPTOS_PREC_INTERNETCONTROL
) :
596 if (ip_options_echo(&icmp_param
->replyopts
.opt
.opt
, skb_in
))
601 * Prepare data for ICMP header.
604 icmp_param
->data
.icmph
.type
= type
;
605 icmp_param
->data
.icmph
.code
= code
;
606 icmp_param
->data
.icmph
.un
.gateway
= info
;
607 icmp_param
->data
.icmph
.checksum
= 0;
608 icmp_param
->skb
= skb_in
;
609 icmp_param
->offset
= skb_network_offset(skb_in
);
610 inet_sk(sk
)->tos
= tos
;
611 ipc
.addr
= iph
->saddr
;
612 ipc
.opt
= &icmp_param
->replyopts
.opt
;
617 rt
= icmp_route_lookup(net
, &fl4
, skb_in
, iph
, saddr
, tos
,
618 type
, code
, icmp_param
);
622 if (!icmpv4_xrlim_allow(net
, rt
, &fl4
, type
, code
))
625 /* RFC says return as much as we can without exceeding 576 bytes. */
627 room
= dst_mtu(&rt
->dst
);
630 room
-= sizeof(struct iphdr
) + icmp_param
->replyopts
.opt
.opt
.optlen
;
631 room
-= sizeof(struct icmphdr
);
633 icmp_param
->data_len
= skb_in
->len
- icmp_param
->offset
;
634 if (icmp_param
->data_len
> room
)
635 icmp_param
->data_len
= room
;
636 icmp_param
->head_len
= sizeof(struct icmphdr
);
638 icmp_push_reply(icmp_param
, &fl4
, &ipc
, &rt
);
642 icmp_xmit_unlock(sk
);
647 EXPORT_SYMBOL(icmp_send
);
650 static void icmp_socket_deliver(struct sk_buff
*skb
, u32 info
)
652 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
653 const struct net_protocol
*ipprot
;
654 int protocol
= iph
->protocol
;
656 /* Checkin full IP header plus 8 bytes of protocol to
657 * avoid additional coding at protocol handlers.
659 if (!pskb_may_pull(skb
, iph
->ihl
* 4 + 8))
662 raw_icmp_error(skb
, protocol
, info
);
665 ipprot
= rcu_dereference(inet_protos
[protocol
]);
666 if (ipprot
&& ipprot
->err_handler
)
667 ipprot
->err_handler(skb
, info
);
672 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, ICMP_QUENCH, and
673 * ICMP_PARAMETERPROB.
676 static void icmp_unreach(struct sk_buff
*skb
)
678 const struct iphdr
*iph
;
679 struct icmphdr
*icmph
;
683 net
= dev_net(skb_dst(skb
)->dev
);
686 * Incomplete header ?
687 * Only checks for the IP header, there should be an
688 * additional check for longer headers in upper levels.
691 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
694 icmph
= icmp_hdr(skb
);
695 iph
= (const struct iphdr
*)skb
->data
;
697 if (iph
->ihl
< 5) /* Mangled header, drop. */
700 if (icmph
->type
== ICMP_DEST_UNREACH
) {
701 switch (icmph
->code
& 15) {
702 case ICMP_NET_UNREACH
:
703 case ICMP_HOST_UNREACH
:
704 case ICMP_PROT_UNREACH
:
705 case ICMP_PORT_UNREACH
:
707 case ICMP_FRAG_NEEDED
:
708 if (ipv4_config
.no_pmtu_disc
) {
709 LIMIT_NETDEBUG(KERN_INFO
pr_fmt("%pI4: fragmentation needed and DF set\n"),
712 info
= ntohs(icmph
->un
.frag
.mtu
);
718 LIMIT_NETDEBUG(KERN_INFO
pr_fmt("%pI4: Source Route Failed\n"),
724 if (icmph
->code
> NR_ICMP_UNREACH
)
726 } else if (icmph
->type
== ICMP_PARAMETERPROB
)
727 info
= ntohl(icmph
->un
.gateway
) >> 24;
730 * Throw it at our lower layers
732 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
734 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
736 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
741 * Check the other end isn't violating RFC 1122. Some routers send
742 * bogus responses to broadcast frames. If you see this message
743 * first check your netmask matches at both ends, if it does then
744 * get the other vendor to fix their kit.
747 if (!net
->ipv4
.sysctl_icmp_ignore_bogus_error_responses
&&
748 inet_addr_type(net
, iph
->daddr
) == RTN_BROADCAST
) {
749 net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
751 icmph
->type
, icmph
->code
,
752 &iph
->daddr
, skb
->dev
->name
);
756 icmp_socket_deliver(skb
, info
);
761 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
767 * Handle ICMP_REDIRECT.
770 static void icmp_redirect(struct sk_buff
*skb
)
772 if (skb
->len
< sizeof(struct iphdr
)) {
773 ICMP_INC_STATS_BH(dev_net(skb
->dev
), ICMP_MIB_INERRORS
);
777 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
780 icmp_socket_deliver(skb
, icmp_hdr(skb
)->un
.gateway
);
784 * Handle ICMP_ECHO ("ping") requests.
786 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
788 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
789 * included in the reply.
790 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
791 * echo requests, MUST have default=NOT.
792 * See also WRT handling of options once they are done and working.
795 static void icmp_echo(struct sk_buff
*skb
)
799 net
= dev_net(skb_dst(skb
)->dev
);
800 if (!net
->ipv4
.sysctl_icmp_echo_ignore_all
) {
801 struct icmp_bxm icmp_param
;
803 icmp_param
.data
.icmph
= *icmp_hdr(skb
);
804 icmp_param
.data
.icmph
.type
= ICMP_ECHOREPLY
;
805 icmp_param
.skb
= skb
;
806 icmp_param
.offset
= 0;
807 icmp_param
.data_len
= skb
->len
;
808 icmp_param
.head_len
= sizeof(struct icmphdr
);
809 icmp_reply(&icmp_param
, skb
);
814 * Handle ICMP Timestamp requests.
815 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
816 * SHOULD be in the kernel for minimum random latency.
817 * MUST be accurate to a few minutes.
818 * MUST be updated at least at 15Hz.
820 static void icmp_timestamp(struct sk_buff
*skb
)
823 struct icmp_bxm icmp_param
;
831 * Fill in the current time as ms since midnight UT:
834 icmp_param
.data
.times
[1] = htonl((tv
.tv_sec
% 86400) * MSEC_PER_SEC
+
835 tv
.tv_nsec
/ NSEC_PER_MSEC
);
836 icmp_param
.data
.times
[2] = icmp_param
.data
.times
[1];
837 if (skb_copy_bits(skb
, 0, &icmp_param
.data
.times
[0], 4))
839 icmp_param
.data
.icmph
= *icmp_hdr(skb
);
840 icmp_param
.data
.icmph
.type
= ICMP_TIMESTAMPREPLY
;
841 icmp_param
.data
.icmph
.code
= 0;
842 icmp_param
.skb
= skb
;
843 icmp_param
.offset
= 0;
844 icmp_param
.data_len
= 0;
845 icmp_param
.head_len
= sizeof(struct icmphdr
) + 12;
846 icmp_reply(&icmp_param
, skb
);
850 ICMP_INC_STATS_BH(dev_net(skb_dst(skb
)->dev
), ICMP_MIB_INERRORS
);
854 static void icmp_discard(struct sk_buff
*skb
)
859 * Deal with incoming ICMP packets.
861 int icmp_rcv(struct sk_buff
*skb
)
863 struct icmphdr
*icmph
;
864 struct rtable
*rt
= skb_rtable(skb
);
865 struct net
*net
= dev_net(rt
->dst
.dev
);
867 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
868 struct sec_path
*sp
= skb_sec_path(skb
);
871 if (!(sp
&& sp
->xvec
[sp
->len
- 1]->props
.flags
&
875 if (!pskb_may_pull(skb
, sizeof(*icmph
) + sizeof(struct iphdr
)))
878 nh
= skb_network_offset(skb
);
879 skb_set_network_header(skb
, sizeof(*icmph
));
881 if (!xfrm4_policy_check_reverse(NULL
, XFRM_POLICY_IN
, skb
))
884 skb_set_network_header(skb
, nh
);
887 ICMP_INC_STATS_BH(net
, ICMP_MIB_INMSGS
);
889 switch (skb
->ip_summed
) {
890 case CHECKSUM_COMPLETE
:
891 if (!csum_fold(skb
->csum
))
896 if (__skb_checksum_complete(skb
))
900 if (!pskb_pull(skb
, sizeof(*icmph
)))
903 icmph
= icmp_hdr(skb
);
905 ICMPMSGIN_INC_STATS_BH(net
, icmph
->type
);
907 * 18 is the highest 'known' ICMP type. Anything else is a mystery
909 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
912 if (icmph
->type
> NR_ICMP_TYPES
)
917 * Parse the ICMP message
920 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
922 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
923 * silently ignored (we let user decide with a sysctl).
924 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
925 * discarded if to broadcast/multicast.
927 if ((icmph
->type
== ICMP_ECHO
||
928 icmph
->type
== ICMP_TIMESTAMP
) &&
929 net
->ipv4
.sysctl_icmp_echo_ignore_broadcasts
) {
932 if (icmph
->type
!= ICMP_ECHO
&&
933 icmph
->type
!= ICMP_TIMESTAMP
&&
934 icmph
->type
!= ICMP_ADDRESS
&&
935 icmph
->type
!= ICMP_ADDRESSREPLY
) {
940 icmp_pointers
[icmph
->type
].handler(skb
);
946 ICMP_INC_STATS_BH(net
, ICMP_MIB_CSUMERRORS
);
948 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
952 void icmp_err(struct sk_buff
*skb
, u32 info
)
954 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
955 int offset
= iph
->ihl
<<2;
956 struct icmphdr
*icmph
= (struct icmphdr
*)(skb
->data
+ offset
);
957 int type
= icmp_hdr(skb
)->type
;
958 int code
= icmp_hdr(skb
)->code
;
959 struct net
*net
= dev_net(skb
->dev
);
962 * Use ping_err to handle all icmp errors except those
963 * triggered by ICMP_ECHOREPLY which sent from kernel.
965 if (icmph
->type
!= ICMP_ECHOREPLY
) {
966 ping_err(skb
, offset
, info
);
970 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
)
971 ipv4_update_pmtu(skb
, net
, info
, 0, 0, IPPROTO_ICMP
, 0);
972 else if (type
== ICMP_REDIRECT
)
973 ipv4_redirect(skb
, net
, 0, 0, IPPROTO_ICMP
, 0);
977 * This table is the definition of how we handle ICMP.
979 static const struct icmp_control icmp_pointers
[NR_ICMP_TYPES
+ 1] = {
984 .handler
= icmp_discard
,
988 .handler
= icmp_discard
,
991 [ICMP_DEST_UNREACH
] = {
992 .handler
= icmp_unreach
,
995 [ICMP_SOURCE_QUENCH
] = {
996 .handler
= icmp_unreach
,
1000 .handler
= icmp_redirect
,
1004 .handler
= icmp_discard
,
1008 .handler
= icmp_discard
,
1012 .handler
= icmp_echo
,
1015 .handler
= icmp_discard
,
1019 .handler
= icmp_discard
,
1022 [ICMP_TIME_EXCEEDED
] = {
1023 .handler
= icmp_unreach
,
1026 [ICMP_PARAMETERPROB
] = {
1027 .handler
= icmp_unreach
,
1030 [ICMP_TIMESTAMP
] = {
1031 .handler
= icmp_timestamp
,
1033 [ICMP_TIMESTAMPREPLY
] = {
1034 .handler
= icmp_discard
,
1036 [ICMP_INFO_REQUEST
] = {
1037 .handler
= icmp_discard
,
1039 [ICMP_INFO_REPLY
] = {
1040 .handler
= icmp_discard
,
1043 .handler
= icmp_discard
,
1045 [ICMP_ADDRESSREPLY
] = {
1046 .handler
= icmp_discard
,
1050 static void __net_exit
icmp_sk_exit(struct net
*net
)
1054 for_each_possible_cpu(i
)
1055 inet_ctl_sock_destroy(net
->ipv4
.icmp_sk
[i
]);
1056 kfree(net
->ipv4
.icmp_sk
);
1057 net
->ipv4
.icmp_sk
= NULL
;
1060 static int __net_init
icmp_sk_init(struct net
*net
)
1065 kzalloc(nr_cpu_ids
* sizeof(struct sock
*), GFP_KERNEL
);
1066 if (net
->ipv4
.icmp_sk
== NULL
)
1069 for_each_possible_cpu(i
) {
1072 err
= inet_ctl_sock_create(&sk
, PF_INET
,
1073 SOCK_RAW
, IPPROTO_ICMP
, net
);
1077 net
->ipv4
.icmp_sk
[i
] = sk
;
1079 /* Enough space for 2 64K ICMP packets, including
1080 * sk_buff/skb_shared_info struct overhead.
1082 sk
->sk_sndbuf
= 2 * SKB_TRUESIZE(64 * 1024);
1085 * Speedup sock_wfree()
1087 sock_set_flag(sk
, SOCK_USE_WRITE_QUEUE
);
1088 inet_sk(sk
)->pmtudisc
= IP_PMTUDISC_DONT
;
1091 /* Control parameters for ECHO replies. */
1092 net
->ipv4
.sysctl_icmp_echo_ignore_all
= 0;
1093 net
->ipv4
.sysctl_icmp_echo_ignore_broadcasts
= 1;
1095 /* Control parameter - ignore bogus broadcast responses? */
1096 net
->ipv4
.sysctl_icmp_ignore_bogus_error_responses
= 1;
1099 * Configurable global rate limit.
1101 * ratelimit defines tokens/packet consumed for dst->rate_token
1102 * bucket ratemask defines which icmp types are ratelimited by
1103 * setting it's bit position.
1106 * dest unreachable (3), source quench (4),
1107 * time exceeded (11), parameter problem (12)
1110 net
->ipv4
.sysctl_icmp_ratelimit
= 1 * HZ
;
1111 net
->ipv4
.sysctl_icmp_ratemask
= 0x1818;
1112 net
->ipv4
.sysctl_icmp_errors_use_inbound_ifaddr
= 0;
1117 for_each_possible_cpu(i
)
1118 inet_ctl_sock_destroy(net
->ipv4
.icmp_sk
[i
]);
1119 kfree(net
->ipv4
.icmp_sk
);
1123 static struct pernet_operations __net_initdata icmp_sk_ops
= {
1124 .init
= icmp_sk_init
,
1125 .exit
= icmp_sk_exit
,
1128 int __init
icmp_init(void)
1130 return register_pernet_subsys(&icmp_sk_ops
);