2 * NET3: Implementation of the ICMP protocol layer.
4 * Alan Cox, <alan@redhat.com>
6 * Version: $Id: icmp.c,v 1.85 2002/02/01 22:01:03 davem Exp $
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
13 * Some of the function names and the icmp unreach table for this
14 * module were derived from [icmp.c 1.0.11 06/02/93] by
15 * Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
16 * Other than that this module is a complete rewrite.
19 * Clemens Fruhwirth : introduce global icmp rate limiting
20 * with icmp type masking ability instead
21 * of broken per type icmp timeouts.
22 * Mike Shaver : RFC1122 checks.
23 * Alan Cox : Multicast ping reply as self.
24 * Alan Cox : Fix atomicity lockup in ip_build_xmit
26 * Alan Cox : Added 216,128 byte paths to the MTU
28 * Martin Mares : RFC1812 checks.
29 * Martin Mares : Can be configured to follow redirects
30 * if acting as a router _without_ a
31 * routing protocol (RFC 1812).
32 * Martin Mares : Echo requests may be configured to
33 * be ignored (RFC 1812).
34 * Martin Mares : Limitation of ICMP error message
35 * transmit rate (RFC 1812).
36 * Martin Mares : TOS and Precedence set correctly
38 * Martin Mares : Now copying as much data from the
39 * original packet as we can without
40 * exceeding 576 bytes (RFC 1812).
41 * Willy Konynenberg : Transparent proxying support.
42 * Keith Owens : RFC1191 correction for 4.2BSD based
44 * Thomas Quinot : ICMP Dest Unreach codes up to 15 are
46 * Andi Kleen : Check all packet lengths properly
47 * and moved all kfree_skb() up to
49 * Andi Kleen : Move the rate limit bookkeeping
50 * into the dest entry and use a token
51 * bucket filter (thanks to ANK). Make
52 * the rates sysctl configurable.
53 * Yu Tianli : Fixed two ugly bugs in icmp_send
54 * - IP option length was accounted wrongly
55 * - ICMP header length was not accounted
57 * Tristan Greaves : Added sysctl option to ignore bogus
58 * broadcast responses from broken routers.
62 * - Should use skb_pull() instead of all the manual checking.
63 * This would also greatly simply some upper layer error handlers. --AK
67 #include <linux/config.h>
68 #include <linux/module.h>
69 #include <linux/types.h>
70 #include <linux/jiffies.h>
71 #include <linux/kernel.h>
72 #include <linux/fcntl.h>
73 #include <linux/socket.h>
75 #include <linux/inet.h>
76 #include <linux/netdevice.h>
77 #include <linux/string.h>
78 #include <linux/netfilter_ipv4.h>
81 #include <net/route.h>
82 #include <net/protocol.h>
87 #include <linux/skbuff.h>
89 #include <linux/errno.h>
90 #include <linux/timer.h>
91 #include <linux/init.h>
92 #include <asm/system.h>
93 #include <asm/uaccess.h>
94 #include <net/checksum.h>
97 * Build xmit assembly blocks
106 struct icmphdr icmph
;
110 struct ip_options replyopts
;
111 unsigned char optbuf
[40];
117 DEFINE_SNMP_STAT(struct icmp_mib
, icmp_statistics
) __read_mostly
;
119 /* An array of errno for error messages from dest unreach. */
120 /* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
122 struct icmp_err icmp_err_convert
[] = {
124 .errno
= ENETUNREACH
, /* ICMP_NET_UNREACH */
128 .errno
= EHOSTUNREACH
, /* ICMP_HOST_UNREACH */
132 .errno
= ENOPROTOOPT
/* ICMP_PROT_UNREACH */,
136 .errno
= ECONNREFUSED
, /* ICMP_PORT_UNREACH */
140 .errno
= EMSGSIZE
, /* ICMP_FRAG_NEEDED */
144 .errno
= EOPNOTSUPP
, /* ICMP_SR_FAILED */
148 .errno
= ENETUNREACH
, /* ICMP_NET_UNKNOWN */
152 .errno
= EHOSTDOWN
, /* ICMP_HOST_UNKNOWN */
156 .errno
= ENONET
, /* ICMP_HOST_ISOLATED */
160 .errno
= ENETUNREACH
, /* ICMP_NET_ANO */
164 .errno
= EHOSTUNREACH
, /* ICMP_HOST_ANO */
168 .errno
= ENETUNREACH
, /* ICMP_NET_UNR_TOS */
172 .errno
= EHOSTUNREACH
, /* ICMP_HOST_UNR_TOS */
176 .errno
= EHOSTUNREACH
, /* ICMP_PKT_FILTERED */
180 .errno
= EHOSTUNREACH
, /* ICMP_PREC_VIOLATION */
184 .errno
= EHOSTUNREACH
, /* ICMP_PREC_CUTOFF */
189 /* Control parameters for ECHO replies. */
190 int sysctl_icmp_echo_ignore_all
;
191 int sysctl_icmp_echo_ignore_broadcasts
= 1;
193 /* Control parameter - ignore bogus broadcast responses? */
194 int sysctl_icmp_ignore_bogus_error_responses
;
197 * Configurable global rate limit.
199 * ratelimit defines tokens/packet consumed for dst->rate_token bucket
200 * ratemask defines which icmp types are ratelimited by setting
204 * dest unreachable (3), source quench (4),
205 * time exceeded (11), parameter problem (12)
208 int sysctl_icmp_ratelimit
= 1 * HZ
;
209 int sysctl_icmp_ratemask
= 0x1818;
210 int sysctl_icmp_errors_use_inbound_ifaddr
;
213 * ICMP control array. This specifies what to do with each ICMP.
216 struct icmp_control
{
217 int output_entry
; /* Field for increment on output */
218 int input_entry
; /* Field for increment on input */
219 void (*handler
)(struct sk_buff
*skb
);
220 short error
; /* This ICMP is classed as an error message */
223 static struct icmp_control icmp_pointers
[NR_ICMP_TYPES
+1];
226 * The ICMP socket(s). This is the most convenient way to flow control
227 * our ICMP output as well as maintain a clean interface throughout
228 * all layers. All Socketless IP sends will soon be gone.
230 * On SMP we have one ICMP socket per-cpu.
232 static DEFINE_PER_CPU(struct socket
*, __icmp_socket
) = NULL
;
233 #define icmp_socket __get_cpu_var(__icmp_socket)
235 static __inline__
int icmp_xmit_lock(void)
239 if (unlikely(!spin_trylock(&icmp_socket
->sk
->sk_lock
.slock
))) {
240 /* This can happen if the output path signals a
241 * dst_link_failure() for an outgoing ICMP packet.
249 static void icmp_xmit_unlock(void)
251 spin_unlock_bh(&icmp_socket
->sk
->sk_lock
.slock
);
255 * Send an ICMP frame.
259 * Check transmit rate limitation for given message.
260 * The rate information is held in the destination cache now.
261 * This function is generic and could be used for other purposes
262 * too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
264 * Note that the same dst_entry fields are modified by functions in
265 * route.c too, but these work for packet destinations while xrlim_allow
266 * works for icmp destinations. This means the rate limiting information
267 * for one "ip object" is shared - and these ICMPs are twice limited:
268 * by source and by destination.
270 * RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
271 * SHOULD allow setting of rate limits
273 * Shared between ICMPv4 and ICMPv6.
275 #define XRLIM_BURST_FACTOR 6
276 int xrlim_allow(struct dst_entry
*dst
, int timeout
)
282 dst
->rate_tokens
+= now
- dst
->rate_last
;
283 dst
->rate_last
= now
;
284 if (dst
->rate_tokens
> XRLIM_BURST_FACTOR
* timeout
)
285 dst
->rate_tokens
= XRLIM_BURST_FACTOR
* timeout
;
286 if (dst
->rate_tokens
>= timeout
) {
287 dst
->rate_tokens
-= timeout
;
293 static inline int icmpv4_xrlim_allow(struct rtable
*rt
, int type
, int code
)
295 struct dst_entry
*dst
= &rt
->u
.dst
;
298 if (type
> NR_ICMP_TYPES
)
301 /* Don't limit PMTU discovery. */
302 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
)
305 /* No rate limit on loopback */
306 if (dst
->dev
&& (dst
->dev
->flags
&IFF_LOOPBACK
))
309 /* Limit if icmp type is enabled in ratemask. */
310 if ((1 << type
) & sysctl_icmp_ratemask
)
311 rc
= xrlim_allow(dst
, sysctl_icmp_ratelimit
);
317 * Maintain the counters used in the SNMP statistics for outgoing ICMP
319 static void icmp_out_count(int type
)
321 if (type
<= NR_ICMP_TYPES
) {
322 ICMP_INC_STATS(icmp_pointers
[type
].output_entry
);
323 ICMP_INC_STATS(ICMP_MIB_OUTMSGS
);
328 * Checksum each fragment, and on the first include the headers and final
331 static int icmp_glue_bits(void *from
, char *to
, int offset
, int len
, int odd
,
334 struct icmp_bxm
*icmp_param
= (struct icmp_bxm
*)from
;
337 csum
= skb_copy_and_csum_bits(icmp_param
->skb
,
338 icmp_param
->offset
+ offset
,
341 skb
->csum
= csum_block_add(skb
->csum
, csum
, odd
);
342 if (icmp_pointers
[icmp_param
->data
.icmph
.type
].error
)
343 nf_ct_attach(skb
, icmp_param
->skb
);
347 static void icmp_push_reply(struct icmp_bxm
*icmp_param
,
348 struct ipcm_cookie
*ipc
, struct rtable
*rt
)
352 if (ip_append_data(icmp_socket
->sk
, icmp_glue_bits
, icmp_param
,
353 icmp_param
->data_len
+icmp_param
->head_len
,
354 icmp_param
->head_len
,
355 ipc
, rt
, MSG_DONTWAIT
) < 0)
356 ip_flush_pending_frames(icmp_socket
->sk
);
357 else if ((skb
= skb_peek(&icmp_socket
->sk
->sk_write_queue
)) != NULL
) {
358 struct icmphdr
*icmph
= skb
->h
.icmph
;
359 unsigned int csum
= 0;
360 struct sk_buff
*skb1
;
362 skb_queue_walk(&icmp_socket
->sk
->sk_write_queue
, skb1
) {
363 csum
= csum_add(csum
, skb1
->csum
);
365 csum
= csum_partial_copy_nocheck((void *)&icmp_param
->data
,
367 icmp_param
->head_len
, csum
);
368 icmph
->checksum
= csum_fold(csum
);
369 skb
->ip_summed
= CHECKSUM_NONE
;
370 ip_push_pending_frames(icmp_socket
->sk
);
375 * Driving logic for building and sending ICMP messages.
378 static void icmp_reply(struct icmp_bxm
*icmp_param
, struct sk_buff
*skb
)
380 struct sock
*sk
= icmp_socket
->sk
;
381 struct inet_sock
*inet
= inet_sk(sk
);
382 struct ipcm_cookie ipc
;
383 struct rtable
*rt
= (struct rtable
*)skb
->dst
;
386 if (ip_options_echo(&icmp_param
->replyopts
, skb
))
389 if (icmp_xmit_lock())
392 icmp_param
->data
.icmph
.checksum
= 0;
393 icmp_out_count(icmp_param
->data
.icmph
.type
);
395 inet
->tos
= skb
->nh
.iph
->tos
;
396 daddr
= ipc
.addr
= rt
->rt_src
;
398 if (icmp_param
->replyopts
.optlen
) {
399 ipc
.opt
= &icmp_param
->replyopts
;
401 daddr
= icmp_param
->replyopts
.faddr
;
404 struct flowi fl
= { .nl_u
= { .ip4_u
=
406 .saddr
= rt
->rt_spec_dst
,
407 .tos
= RT_TOS(skb
->nh
.iph
->tos
) } },
408 .proto
= IPPROTO_ICMP
};
409 if (ip_route_output_key(&rt
, &fl
))
412 if (icmpv4_xrlim_allow(rt
, icmp_param
->data
.icmph
.type
,
413 icmp_param
->data
.icmph
.code
))
414 icmp_push_reply(icmp_param
, &ipc
, rt
);
423 * Send an ICMP message in response to a situation
425 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
426 * MAY send more (we do).
427 * MUST NOT change this header information.
428 * MUST NOT reply to a multicast/broadcast IP address.
429 * MUST NOT reply to a multicast/broadcast MAC address.
430 * MUST reply to only the first fragment.
433 void icmp_send(struct sk_buff
*skb_in
, int type
, int code
, u32 info
)
437 struct icmp_bxm icmp_param
;
438 struct rtable
*rt
= (struct rtable
*)skb_in
->dst
;
439 struct ipcm_cookie ipc
;
447 * Find the original header. It is expected to be valid, of course.
448 * Check this, icmp_send is called from the most obscure devices
451 iph
= skb_in
->nh
.iph
;
453 if ((u8
*)iph
< skb_in
->head
|| (u8
*)(iph
+ 1) > skb_in
->tail
)
457 * No replies to physical multicast/broadcast
459 if (skb_in
->pkt_type
!= PACKET_HOST
)
463 * Now check at the protocol level
465 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))
469 * Only reply to fragment 0. We byte re-order the constant
470 * mask for efficiency.
472 if (iph
->frag_off
& htons(IP_OFFSET
))
476 * If we send an ICMP error to an ICMP error a mess would result..
478 if (icmp_pointers
[type
].error
) {
480 * We are an error, check if we are replying to an
483 if (iph
->protocol
== IPPROTO_ICMP
) {
484 u8 _inner_type
, *itp
;
486 itp
= skb_header_pointer(skb_in
,
489 offsetof(struct icmphdr
,
498 * Assume any unknown ICMP type is an error. This
499 * isn't specified by the RFC, but think about it..
501 if (*itp
> NR_ICMP_TYPES
||
502 icmp_pointers
[*itp
].error
)
507 if (icmp_xmit_lock())
511 * Construct source address and options.
515 if (!(rt
->rt_flags
& RTCF_LOCAL
)) {
516 if (sysctl_icmp_errors_use_inbound_ifaddr
)
517 saddr
= inet_select_addr(skb_in
->dev
, 0, RT_SCOPE_LINK
);
522 tos
= icmp_pointers
[type
].error
? ((iph
->tos
& IPTOS_TOS_MASK
) |
523 IPTOS_PREC_INTERNETCONTROL
) :
526 if (ip_options_echo(&icmp_param
.replyopts
, skb_in
))
531 * Prepare data for ICMP header.
534 icmp_param
.data
.icmph
.type
= type
;
535 icmp_param
.data
.icmph
.code
= code
;
536 icmp_param
.data
.icmph
.un
.gateway
= info
;
537 icmp_param
.data
.icmph
.checksum
= 0;
538 icmp_param
.skb
= skb_in
;
539 icmp_param
.offset
= skb_in
->nh
.raw
- skb_in
->data
;
540 icmp_out_count(icmp_param
.data
.icmph
.type
);
541 inet_sk(icmp_socket
->sk
)->tos
= tos
;
542 ipc
.addr
= iph
->saddr
;
543 ipc
.opt
= &icmp_param
.replyopts
;
549 .daddr
= icmp_param
.replyopts
.srr
?
550 icmp_param
.replyopts
.faddr
:
556 .proto
= IPPROTO_ICMP
,
564 if (ip_route_output_key(&rt
, &fl
))
568 if (!icmpv4_xrlim_allow(rt
, type
, code
))
571 /* RFC says return as much as we can without exceeding 576 bytes. */
573 room
= dst_mtu(&rt
->u
.dst
);
576 room
-= sizeof(struct iphdr
) + icmp_param
.replyopts
.optlen
;
577 room
-= sizeof(struct icmphdr
);
579 icmp_param
.data_len
= skb_in
->len
- icmp_param
.offset
;
580 if (icmp_param
.data_len
> room
)
581 icmp_param
.data_len
= room
;
582 icmp_param
.head_len
= sizeof(struct icmphdr
);
584 icmp_push_reply(&icmp_param
, &ipc
, rt
);
594 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
597 static void icmp_unreach(struct sk_buff
*skb
)
600 struct icmphdr
*icmph
;
602 struct net_protocol
*ipprot
;
607 * Incomplete header ?
608 * Only checks for the IP header, there should be an
609 * additional check for longer headers in upper levels.
612 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
615 icmph
= skb
->h
.icmph
;
616 iph
= (struct iphdr
*)skb
->data
;
618 if (iph
->ihl
< 5) /* Mangled header, drop. */
621 if (icmph
->type
== ICMP_DEST_UNREACH
) {
622 switch (icmph
->code
& 15) {
623 case ICMP_NET_UNREACH
:
624 case ICMP_HOST_UNREACH
:
625 case ICMP_PROT_UNREACH
:
626 case ICMP_PORT_UNREACH
:
628 case ICMP_FRAG_NEEDED
:
629 if (ipv4_config
.no_pmtu_disc
) {
630 LIMIT_NETDEBUG(KERN_INFO
"ICMP: %u.%u.%u.%u: "
631 "fragmentation needed "
633 NIPQUAD(iph
->daddr
));
635 info
= ip_rt_frag_needed(iph
,
636 ntohs(icmph
->un
.frag
.mtu
));
642 LIMIT_NETDEBUG(KERN_INFO
"ICMP: %u.%u.%u.%u: Source "
644 NIPQUAD(iph
->daddr
));
649 if (icmph
->code
> NR_ICMP_UNREACH
)
651 } else if (icmph
->type
== ICMP_PARAMETERPROB
)
652 info
= ntohl(icmph
->un
.gateway
) >> 24;
655 * Throw it at our lower layers
657 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
659 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
661 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
666 * Check the other end isnt violating RFC 1122. Some routers send
667 * bogus responses to broadcast frames. If you see this message
668 * first check your netmask matches at both ends, if it does then
669 * get the other vendor to fix their kit.
672 if (!sysctl_icmp_ignore_bogus_error_responses
&&
673 inet_addr_type(iph
->daddr
) == RTN_BROADCAST
) {
675 printk(KERN_WARNING
"%u.%u.%u.%u sent an invalid ICMP "
677 "error to a broadcast: %u.%u.%u.%u on %s\n",
678 NIPQUAD(skb
->nh
.iph
->saddr
),
679 icmph
->type
, icmph
->code
,
685 /* Checkin full IP header plus 8 bytes of protocol to
686 * avoid additional coding at protocol handlers.
688 if (!pskb_may_pull(skb
, iph
->ihl
* 4 + 8))
691 iph
= (struct iphdr
*)skb
->data
;
692 protocol
= iph
->protocol
;
695 * Deliver ICMP message to raw sockets. Pretty useless feature?
698 /* Note: See raw.c and net/raw.h, RAWV4_HTABLE_SIZE==MAX_INET_PROTOS */
699 hash
= protocol
& (MAX_INET_PROTOS
- 1);
700 read_lock(&raw_v4_lock
);
701 if ((raw_sk
= sk_head(&raw_v4_htable
[hash
])) != NULL
) {
702 while ((raw_sk
= __raw_v4_lookup(raw_sk
, protocol
, iph
->daddr
,
704 skb
->dev
->ifindex
)) != NULL
) {
705 raw_err(raw_sk
, skb
, info
);
706 raw_sk
= sk_next(raw_sk
);
707 iph
= (struct iphdr
*)skb
->data
;
710 read_unlock(&raw_v4_lock
);
713 ipprot
= rcu_dereference(inet_protos
[hash
]);
714 if (ipprot
&& ipprot
->err_handler
)
715 ipprot
->err_handler(skb
, info
);
721 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS
);
727 * Handle ICMP_REDIRECT.
730 static void icmp_redirect(struct sk_buff
*skb
)
735 if (skb
->len
< sizeof(struct iphdr
))
739 * Get the copied header of the packet that caused the redirect
741 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
744 iph
= (struct iphdr
*)skb
->data
;
747 switch (skb
->h
.icmph
->code
& 7) {
749 case ICMP_REDIR_NETTOS
:
751 * As per RFC recommendations now handle it as a host redirect.
753 case ICMP_REDIR_HOST
:
754 case ICMP_REDIR_HOSTTOS
:
755 ip_rt_redirect(skb
->nh
.iph
->saddr
, ip
, skb
->h
.icmph
->un
.gateway
,
756 iph
->saddr
, iph
->tos
, skb
->dev
);
762 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS
);
767 * Handle ICMP_ECHO ("ping") requests.
769 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
771 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
772 * included in the reply.
773 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
774 * echo requests, MUST have default=NOT.
775 * See also WRT handling of options once they are done and working.
778 static void icmp_echo(struct sk_buff
*skb
)
780 if (!sysctl_icmp_echo_ignore_all
) {
781 struct icmp_bxm icmp_param
;
783 icmp_param
.data
.icmph
= *skb
->h
.icmph
;
784 icmp_param
.data
.icmph
.type
= ICMP_ECHOREPLY
;
785 icmp_param
.skb
= skb
;
786 icmp_param
.offset
= 0;
787 icmp_param
.data_len
= skb
->len
;
788 icmp_param
.head_len
= sizeof(struct icmphdr
);
789 icmp_reply(&icmp_param
, skb
);
794 * Handle ICMP Timestamp requests.
795 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
796 * SHOULD be in the kernel for minimum random latency.
797 * MUST be accurate to a few minutes.
798 * MUST be updated at least at 15Hz.
800 static void icmp_timestamp(struct sk_buff
*skb
)
803 struct icmp_bxm icmp_param
;
811 * Fill in the current time as ms since midnight UT:
813 do_gettimeofday(&tv
);
814 icmp_param
.data
.times
[1] = htonl((tv
.tv_sec
% 86400) * 1000 +
816 icmp_param
.data
.times
[2] = icmp_param
.data
.times
[1];
817 if (skb_copy_bits(skb
, 0, &icmp_param
.data
.times
[0], 4))
819 icmp_param
.data
.icmph
= *skb
->h
.icmph
;
820 icmp_param
.data
.icmph
.type
= ICMP_TIMESTAMPREPLY
;
821 icmp_param
.data
.icmph
.code
= 0;
822 icmp_param
.skb
= skb
;
823 icmp_param
.offset
= 0;
824 icmp_param
.data_len
= 0;
825 icmp_param
.head_len
= sizeof(struct icmphdr
) + 12;
826 icmp_reply(&icmp_param
, skb
);
830 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS
);
836 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
838 * RFC1122 (3.2.2.9). A host MUST only send replies to
839 * ADDRESS_MASK requests if it's been configured as an address mask
840 * agent. Receiving a request doesn't constitute implicit permission to
841 * act as one. Of course, implementing this correctly requires (SHOULD)
842 * a way to turn the functionality on and off. Another one for sysctl(),
845 * RFC1812 (4.3.3.9). A router MUST implement it.
846 * A router SHOULD have switch turning it on/off.
847 * This switch MUST be ON by default.
849 * Gratuitous replies, zero-source replies are not implemented,
850 * that complies with RFC. DO NOT implement them!!! All the idea
851 * of broadcast addrmask replies as specified in RFC950 is broken.
852 * The problem is that it is not uncommon to have several prefixes
853 * on one physical interface. Moreover, addrmask agent can even be
854 * not aware of existing another prefixes.
855 * If source is zero, addrmask agent cannot choose correct prefix.
856 * Gratuitous mask announcements suffer from the same problem.
857 * RFC1812 explains it, but still allows to use ADDRMASK,
858 * that is pretty silly. --ANK
860 * All these rules are so bizarre, that I removed kernel addrmask
861 * support at all. It is wrong, it is obsolete, nobody uses it in
864 * Furthermore you can do it with a usermode address agent program
868 static void icmp_address(struct sk_buff
*skb
)
872 printk(KERN_DEBUG
"a guy asks for address mask. Who is it?\n");
877 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
878 * loudly if an inconsistency is found.
881 static void icmp_address_reply(struct sk_buff
*skb
)
883 struct rtable
*rt
= (struct rtable
*)skb
->dst
;
884 struct net_device
*dev
= skb
->dev
;
885 struct in_device
*in_dev
;
886 struct in_ifaddr
*ifa
;
888 if (skb
->len
< 4 || !(rt
->rt_flags
&RTCF_DIRECTSRC
))
891 in_dev
= in_dev_get(dev
);
895 if (in_dev
->ifa_list
&&
896 IN_DEV_LOG_MARTIANS(in_dev
) &&
897 IN_DEV_FORWARD(in_dev
)) {
900 mp
= skb_header_pointer(skb
, 0, sizeof(_mask
), &_mask
);
903 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
) {
904 if (*mp
== ifa
->ifa_mask
&&
905 inet_ifa_match(rt
->rt_src
, ifa
))
908 if (!ifa
&& net_ratelimit()) {
909 printk(KERN_INFO
"Wrong address mask %u.%u.%u.%u from "
911 NIPQUAD(*mp
), dev
->name
, NIPQUAD(rt
->rt_src
));
919 static void icmp_discard(struct sk_buff
*skb
)
924 * Deal with incoming ICMP packets.
926 int icmp_rcv(struct sk_buff
*skb
)
928 struct icmphdr
*icmph
;
929 struct rtable
*rt
= (struct rtable
*)skb
->dst
;
931 ICMP_INC_STATS_BH(ICMP_MIB_INMSGS
);
933 switch (skb
->ip_summed
) {
935 if (!(u16
)csum_fold(skb
->csum
))
937 LIMIT_NETDEBUG(KERN_DEBUG
"icmp v4 hw csum failure\n");
939 if ((u16
)csum_fold(skb_checksum(skb
, 0, skb
->len
, 0)))
944 if (!pskb_pull(skb
, sizeof(struct icmphdr
)))
947 icmph
= skb
->h
.icmph
;
950 * 18 is the highest 'known' ICMP type. Anything else is a mystery
952 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
955 if (icmph
->type
> NR_ICMP_TYPES
)
960 * Parse the ICMP message
963 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
965 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
966 * silently ignored (we let user decide with a sysctl).
967 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
968 * discarded if to broadcast/multicast.
970 if ((icmph
->type
== ICMP_ECHO
||
971 icmph
->type
== ICMP_TIMESTAMP
) &&
972 sysctl_icmp_echo_ignore_broadcasts
) {
975 if (icmph
->type
!= ICMP_ECHO
&&
976 icmph
->type
!= ICMP_TIMESTAMP
&&
977 icmph
->type
!= ICMP_ADDRESS
&&
978 icmph
->type
!= ICMP_ADDRESSREPLY
) {
983 ICMP_INC_STATS_BH(icmp_pointers
[icmph
->type
].input_entry
);
984 icmp_pointers
[icmph
->type
].handler(skb
);
990 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS
);
995 * This table is the definition of how we handle ICMP.
997 static struct icmp_control icmp_pointers
[NR_ICMP_TYPES
+ 1] = {
999 .output_entry
= ICMP_MIB_OUTECHOREPS
,
1000 .input_entry
= ICMP_MIB_INECHOREPS
,
1001 .handler
= icmp_discard
,
1004 .output_entry
= ICMP_MIB_DUMMY
,
1005 .input_entry
= ICMP_MIB_INERRORS
,
1006 .handler
= icmp_discard
,
1010 .output_entry
= ICMP_MIB_DUMMY
,
1011 .input_entry
= ICMP_MIB_INERRORS
,
1012 .handler
= icmp_discard
,
1015 [ICMP_DEST_UNREACH
] = {
1016 .output_entry
= ICMP_MIB_OUTDESTUNREACHS
,
1017 .input_entry
= ICMP_MIB_INDESTUNREACHS
,
1018 .handler
= icmp_unreach
,
1021 [ICMP_SOURCE_QUENCH
] = {
1022 .output_entry
= ICMP_MIB_OUTSRCQUENCHS
,
1023 .input_entry
= ICMP_MIB_INSRCQUENCHS
,
1024 .handler
= icmp_unreach
,
1028 .output_entry
= ICMP_MIB_OUTREDIRECTS
,
1029 .input_entry
= ICMP_MIB_INREDIRECTS
,
1030 .handler
= icmp_redirect
,
1034 .output_entry
= ICMP_MIB_DUMMY
,
1035 .input_entry
= ICMP_MIB_INERRORS
,
1036 .handler
= icmp_discard
,
1040 .output_entry
= ICMP_MIB_DUMMY
,
1041 .input_entry
= ICMP_MIB_INERRORS
,
1042 .handler
= icmp_discard
,
1046 .output_entry
= ICMP_MIB_OUTECHOS
,
1047 .input_entry
= ICMP_MIB_INECHOS
,
1048 .handler
= icmp_echo
,
1051 .output_entry
= ICMP_MIB_DUMMY
,
1052 .input_entry
= ICMP_MIB_INERRORS
,
1053 .handler
= icmp_discard
,
1057 .output_entry
= ICMP_MIB_DUMMY
,
1058 .input_entry
= ICMP_MIB_INERRORS
,
1059 .handler
= icmp_discard
,
1062 [ICMP_TIME_EXCEEDED
] = {
1063 .output_entry
= ICMP_MIB_OUTTIMEEXCDS
,
1064 .input_entry
= ICMP_MIB_INTIMEEXCDS
,
1065 .handler
= icmp_unreach
,
1068 [ICMP_PARAMETERPROB
] = {
1069 .output_entry
= ICMP_MIB_OUTPARMPROBS
,
1070 .input_entry
= ICMP_MIB_INPARMPROBS
,
1071 .handler
= icmp_unreach
,
1074 [ICMP_TIMESTAMP
] = {
1075 .output_entry
= ICMP_MIB_OUTTIMESTAMPS
,
1076 .input_entry
= ICMP_MIB_INTIMESTAMPS
,
1077 .handler
= icmp_timestamp
,
1079 [ICMP_TIMESTAMPREPLY
] = {
1080 .output_entry
= ICMP_MIB_OUTTIMESTAMPREPS
,
1081 .input_entry
= ICMP_MIB_INTIMESTAMPREPS
,
1082 .handler
= icmp_discard
,
1084 [ICMP_INFO_REQUEST
] = {
1085 .output_entry
= ICMP_MIB_DUMMY
,
1086 .input_entry
= ICMP_MIB_DUMMY
,
1087 .handler
= icmp_discard
,
1089 [ICMP_INFO_REPLY
] = {
1090 .output_entry
= ICMP_MIB_DUMMY
,
1091 .input_entry
= ICMP_MIB_DUMMY
,
1092 .handler
= icmp_discard
,
1095 .output_entry
= ICMP_MIB_OUTADDRMASKS
,
1096 .input_entry
= ICMP_MIB_INADDRMASKS
,
1097 .handler
= icmp_address
,
1099 [ICMP_ADDRESSREPLY
] = {
1100 .output_entry
= ICMP_MIB_OUTADDRMASKREPS
,
1101 .input_entry
= ICMP_MIB_INADDRMASKREPS
,
1102 .handler
= icmp_address_reply
,
1106 void __init
icmp_init(struct net_proto_family
*ops
)
1108 struct inet_sock
*inet
;
1111 for (i
= 0; i
< NR_CPUS
; i
++) {
1114 if (!cpu_possible(i
))
1117 err
= sock_create_kern(PF_INET
, SOCK_RAW
, IPPROTO_ICMP
,
1118 &per_cpu(__icmp_socket
, i
));
1121 panic("Failed to create the ICMP control socket.\n");
1123 per_cpu(__icmp_socket
, i
)->sk
->sk_allocation
= GFP_ATOMIC
;
1125 /* Enough space for 2 64K ICMP packets, including
1126 * sk_buff struct overhead.
1128 per_cpu(__icmp_socket
, i
)->sk
->sk_sndbuf
=
1129 (2 * ((64 * 1024) + sizeof(struct sk_buff
)));
1131 inet
= inet_sk(per_cpu(__icmp_socket
, i
)->sk
);
1133 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
1135 /* Unhash it so that IP input processing does not even
1136 * see it, we do not wish this socket to see incoming
1139 per_cpu(__icmp_socket
, i
)->sk
->sk_prot
->unhash(per_cpu(__icmp_socket
, i
)->sk
);
1143 EXPORT_SYMBOL(icmp_err_convert
);
1144 EXPORT_SYMBOL(icmp_send
);
1145 EXPORT_SYMBOL(icmp_statistics
);
1146 EXPORT_SYMBOL(xrlim_allow
);