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 #include <linux/module.h>
66 #include <linux/types.h>
67 #include <linux/jiffies.h>
68 #include <linux/kernel.h>
69 #include <linux/fcntl.h>
70 #include <linux/socket.h>
72 #include <linux/inet.h>
73 #include <linux/inetdevice.h>
74 #include <linux/netdevice.h>
75 #include <linux/string.h>
76 #include <linux/netfilter_ipv4.h>
77 #include <linux/slab.h>
80 #include <net/route.h>
81 #include <net/protocol.h>
86 #include <linux/skbuff.h>
88 #include <linux/errno.h>
89 #include <linux/timer.h>
90 #include <linux/init.h>
91 #include <asm/system.h>
92 #include <asm/uaccess.h>
93 #include <net/checksum.h>
95 #include <net/inet_common.h>
98 * Build xmit assembly blocks
107 struct icmphdr icmph
;
111 struct ip_options replyopts
;
112 unsigned char optbuf
[40];
115 /* An array of errno for error messages from dest unreach. */
116 /* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
118 const struct icmp_err icmp_err_convert
[] = {
120 .errno
= ENETUNREACH
, /* ICMP_NET_UNREACH */
124 .errno
= EHOSTUNREACH
, /* ICMP_HOST_UNREACH */
128 .errno
= ENOPROTOOPT
/* ICMP_PROT_UNREACH */,
132 .errno
= ECONNREFUSED
, /* ICMP_PORT_UNREACH */
136 .errno
= EMSGSIZE
, /* ICMP_FRAG_NEEDED */
140 .errno
= EOPNOTSUPP
, /* ICMP_SR_FAILED */
144 .errno
= ENETUNREACH
, /* ICMP_NET_UNKNOWN */
148 .errno
= EHOSTDOWN
, /* ICMP_HOST_UNKNOWN */
152 .errno
= ENONET
, /* ICMP_HOST_ISOLATED */
156 .errno
= ENETUNREACH
, /* ICMP_NET_ANO */
160 .errno
= EHOSTUNREACH
, /* ICMP_HOST_ANO */
164 .errno
= ENETUNREACH
, /* ICMP_NET_UNR_TOS */
168 .errno
= EHOSTUNREACH
, /* ICMP_HOST_UNR_TOS */
172 .errno
= EHOSTUNREACH
, /* ICMP_PKT_FILTERED */
176 .errno
= EHOSTUNREACH
, /* ICMP_PREC_VIOLATION */
180 .errno
= EHOSTUNREACH
, /* ICMP_PREC_CUTOFF */
184 EXPORT_SYMBOL(icmp_err_convert
);
187 * ICMP control array. This specifies what to do with each ICMP.
190 struct icmp_control
{
191 void (*handler
)(struct sk_buff
*skb
);
192 short error
; /* This ICMP is classed as an error message */
195 static const struct icmp_control icmp_pointers
[NR_ICMP_TYPES
+1];
198 * The ICMP socket(s). This is the most convenient way to flow control
199 * our ICMP output as well as maintain a clean interface throughout
200 * all layers. All Socketless IP sends will soon be gone.
202 * On SMP we have one ICMP socket per-cpu.
204 static struct sock
*icmp_sk(struct net
*net
)
206 return net
->ipv4
.icmp_sk
[smp_processor_id()];
209 static inline struct sock
*icmp_xmit_lock(struct net
*net
)
217 if (unlikely(!spin_trylock(&sk
->sk_lock
.slock
))) {
218 /* This can happen if the output path signals a
219 * dst_link_failure() for an outgoing ICMP packet.
227 static inline void icmp_xmit_unlock(struct sock
*sk
)
229 spin_unlock_bh(&sk
->sk_lock
.slock
);
233 * Send an ICMP frame.
236 static inline bool icmpv4_xrlim_allow(struct net
*net
, struct rtable
*rt
,
239 struct dst_entry
*dst
= &rt
->dst
;
242 if (type
> NR_ICMP_TYPES
)
245 /* Don't limit PMTU discovery. */
246 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
)
249 /* No rate limit on loopback */
250 if (dst
->dev
&& (dst
->dev
->flags
&IFF_LOOPBACK
))
253 /* Limit if icmp type is enabled in ratemask. */
254 if ((1 << type
) & net
->ipv4
.sysctl_icmp_ratemask
) {
257 rc
= inet_peer_xrlim_allow(rt
->peer
,
258 net
->ipv4
.sysctl_icmp_ratelimit
);
265 * Maintain the counters used in the SNMP statistics for outgoing ICMP
267 void icmp_out_count(struct net
*net
, unsigned char type
)
269 ICMPMSGOUT_INC_STATS(net
, type
);
270 ICMP_INC_STATS(net
, ICMP_MIB_OUTMSGS
);
274 * Checksum each fragment, and on the first include the headers and final
277 static int icmp_glue_bits(void *from
, char *to
, int offset
, int len
, int odd
,
280 struct icmp_bxm
*icmp_param
= (struct icmp_bxm
*)from
;
283 csum
= skb_copy_and_csum_bits(icmp_param
->skb
,
284 icmp_param
->offset
+ offset
,
287 skb
->csum
= csum_block_add(skb
->csum
, csum
, odd
);
288 if (icmp_pointers
[icmp_param
->data
.icmph
.type
].error
)
289 nf_ct_attach(skb
, icmp_param
->skb
);
293 static void icmp_push_reply(struct icmp_bxm
*icmp_param
,
294 struct ipcm_cookie
*ipc
, struct rtable
**rt
)
299 sk
= icmp_sk(dev_net((*rt
)->dst
.dev
));
300 if (ip_append_data(sk
, icmp_glue_bits
, icmp_param
,
301 icmp_param
->data_len
+icmp_param
->head_len
,
302 icmp_param
->head_len
,
303 ipc
, rt
, MSG_DONTWAIT
) < 0) {
304 ICMP_INC_STATS_BH(sock_net(sk
), ICMP_MIB_OUTERRORS
);
305 ip_flush_pending_frames(sk
);
306 } else if ((skb
= skb_peek(&sk
->sk_write_queue
)) != NULL
) {
307 struct icmphdr
*icmph
= icmp_hdr(skb
);
309 struct sk_buff
*skb1
;
311 skb_queue_walk(&sk
->sk_write_queue
, skb1
) {
312 csum
= csum_add(csum
, skb1
->csum
);
314 csum
= csum_partial_copy_nocheck((void *)&icmp_param
->data
,
316 icmp_param
->head_len
, csum
);
317 icmph
->checksum
= csum_fold(csum
);
318 skb
->ip_summed
= CHECKSUM_NONE
;
319 ip_push_pending_frames(sk
);
324 * Driving logic for building and sending ICMP messages.
327 static void icmp_reply(struct icmp_bxm
*icmp_param
, struct sk_buff
*skb
)
329 struct ipcm_cookie ipc
;
330 struct rtable
*rt
= skb_rtable(skb
);
331 struct net
*net
= dev_net(rt
->dst
.dev
);
333 struct inet_sock
*inet
;
336 if (ip_options_echo(&icmp_param
->replyopts
, skb
))
339 sk
= icmp_xmit_lock(net
);
344 icmp_param
->data
.icmph
.checksum
= 0;
346 inet
->tos
= ip_hdr(skb
)->tos
;
347 daddr
= ipc
.addr
= rt
->rt_src
;
350 if (icmp_param
->replyopts
.optlen
) {
351 ipc
.opt
= &icmp_param
->replyopts
;
353 daddr
= icmp_param
->replyopts
.faddr
;
356 struct flowi fl
= { .fl4_dst
= daddr
,
357 .fl4_src
= rt
->rt_spec_dst
,
358 .fl4_tos
= RT_TOS(ip_hdr(skb
)->tos
),
359 .proto
= IPPROTO_ICMP
};
360 security_skb_classify_flow(skb
, &fl
);
361 rt
= ip_route_output_key(net
, &fl
);
365 if (icmpv4_xrlim_allow(net
, rt
, icmp_param
->data
.icmph
.type
,
366 icmp_param
->data
.icmph
.code
))
367 icmp_push_reply(icmp_param
, &ipc
, &rt
);
370 icmp_xmit_unlock(sk
);
373 static struct rtable
*icmp_route_lookup(struct net
*net
, struct sk_buff
*skb_in
,
375 __be32 saddr
, u8 tos
,
377 struct icmp_bxm
*param
)
380 .fl4_dst
= (param
->replyopts
.srr
?
381 param
->replyopts
.faddr
: iph
->saddr
),
383 .fl4_tos
= RT_TOS(tos
),
384 .proto
= IPPROTO_ICMP
,
385 .fl_icmp_type
= type
,
386 .fl_icmp_code
= code
,
388 struct rtable
*rt
, *rt2
;
391 security_skb_classify_flow(skb_in
, &fl
);
392 rt
= __ip_route_output_key(net
, &fl
);
396 /* No need to clone since we're just using its address. */
400 fl
.fl4_src
= rt
->rt_src
;
402 rt
= (struct rtable
*) xfrm_lookup(net
, &rt
->dst
, &fl
, NULL
, 0);
406 } else if (PTR_ERR(rt
) == -EPERM
) {
411 err
= xfrm_decode_session_reverse(skb_in
, &fl
, AF_INET
);
413 goto relookup_failed
;
415 if (inet_addr_type(net
, fl
.fl4_src
) == RTN_LOCAL
) {
416 rt2
= __ip_route_output_key(net
, &fl
);
420 struct flowi fl2
= {};
421 unsigned long orefdst
;
423 fl2
.fl4_dst
= fl
.fl4_src
;
424 rt2
= ip_route_output_key(net
, &fl2
);
427 goto relookup_failed
;
430 orefdst
= skb_in
->_skb_refdst
; /* save old refdst */
431 err
= ip_route_input(skb_in
, fl
.fl4_dst
, fl
.fl4_src
,
432 RT_TOS(tos
), rt2
->dst
.dev
);
434 dst_release(&rt2
->dst
);
435 rt2
= skb_rtable(skb_in
);
436 skb_in
->_skb_refdst
= orefdst
; /* restore old refdst */
440 goto relookup_failed
;
442 rt2
= (struct rtable
*) xfrm_lookup(net
, &rt2
->dst
, &fl
, NULL
, XFRM_LOOKUP_ICMP
);
444 dst_release(&rt
->dst
);
446 } else if (PTR_ERR(rt2
) == -EPERM
) {
448 dst_release(&rt
->dst
);
452 goto relookup_failed
;
463 * Send an ICMP message in response to a situation
465 * RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
466 * MAY send more (we do).
467 * MUST NOT change this header information.
468 * MUST NOT reply to a multicast/broadcast IP address.
469 * MUST NOT reply to a multicast/broadcast MAC address.
470 * MUST reply to only the first fragment.
473 void icmp_send(struct sk_buff
*skb_in
, int type
, int code
, __be32 info
)
477 struct icmp_bxm icmp_param
;
478 struct rtable
*rt
= skb_rtable(skb_in
);
479 struct ipcm_cookie ipc
;
487 net
= dev_net(rt
->dst
.dev
);
490 * Find the original header. It is expected to be valid, of course.
491 * Check this, icmp_send is called from the most obscure devices
494 iph
= ip_hdr(skb_in
);
496 if ((u8
*)iph
< skb_in
->head
||
497 (skb_in
->network_header
+ sizeof(*iph
)) > skb_in
->tail
)
501 * No replies to physical multicast/broadcast
503 if (skb_in
->pkt_type
!= PACKET_HOST
)
507 * Now check at the protocol level
509 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))
513 * Only reply to fragment 0. We byte re-order the constant
514 * mask for efficiency.
516 if (iph
->frag_off
& htons(IP_OFFSET
))
520 * If we send an ICMP error to an ICMP error a mess would result..
522 if (icmp_pointers
[type
].error
) {
524 * We are an error, check if we are replying to an
527 if (iph
->protocol
== IPPROTO_ICMP
) {
528 u8 _inner_type
, *itp
;
530 itp
= skb_header_pointer(skb_in
,
531 skb_network_header(skb_in
) +
533 offsetof(struct icmphdr
,
542 * Assume any unknown ICMP type is an error. This
543 * isn't specified by the RFC, but think about it..
545 if (*itp
> NR_ICMP_TYPES
||
546 icmp_pointers
[*itp
].error
)
551 sk
= icmp_xmit_lock(net
);
556 * Construct source address and options.
560 if (!(rt
->rt_flags
& RTCF_LOCAL
)) {
561 struct net_device
*dev
= NULL
;
564 if (rt_is_input_route(rt
) &&
565 net
->ipv4
.sysctl_icmp_errors_use_inbound_ifaddr
)
566 dev
= dev_get_by_index_rcu(net
, rt
->rt_iif
);
569 saddr
= inet_select_addr(dev
, 0, RT_SCOPE_LINK
);
575 tos
= icmp_pointers
[type
].error
? ((iph
->tos
& IPTOS_TOS_MASK
) |
576 IPTOS_PREC_INTERNETCONTROL
) :
579 if (ip_options_echo(&icmp_param
.replyopts
, skb_in
))
584 * Prepare data for ICMP header.
587 icmp_param
.data
.icmph
.type
= type
;
588 icmp_param
.data
.icmph
.code
= code
;
589 icmp_param
.data
.icmph
.un
.gateway
= info
;
590 icmp_param
.data
.icmph
.checksum
= 0;
591 icmp_param
.skb
= skb_in
;
592 icmp_param
.offset
= skb_network_offset(skb_in
);
593 inet_sk(sk
)->tos
= tos
;
594 ipc
.addr
= iph
->saddr
;
595 ipc
.opt
= &icmp_param
.replyopts
;
598 rt
= icmp_route_lookup(net
, skb_in
, iph
, saddr
, tos
,
599 type
, code
, &icmp_param
);
603 if (!icmpv4_xrlim_allow(net
, rt
, type
, code
))
606 /* RFC says return as much as we can without exceeding 576 bytes. */
608 room
= dst_mtu(&rt
->dst
);
611 room
-= sizeof(struct iphdr
) + icmp_param
.replyopts
.optlen
;
612 room
-= sizeof(struct icmphdr
);
614 icmp_param
.data_len
= skb_in
->len
- icmp_param
.offset
;
615 if (icmp_param
.data_len
> room
)
616 icmp_param
.data_len
= room
;
617 icmp_param
.head_len
= sizeof(struct icmphdr
);
619 icmp_push_reply(&icmp_param
, &ipc
, &rt
);
623 icmp_xmit_unlock(sk
);
626 EXPORT_SYMBOL(icmp_send
);
630 * Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
633 static void icmp_unreach(struct sk_buff
*skb
)
636 struct icmphdr
*icmph
;
638 const struct net_protocol
*ipprot
;
642 net
= dev_net(skb_dst(skb
)->dev
);
645 * Incomplete header ?
646 * Only checks for the IP header, there should be an
647 * additional check for longer headers in upper levels.
650 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
653 icmph
= icmp_hdr(skb
);
654 iph
= (struct iphdr
*)skb
->data
;
656 if (iph
->ihl
< 5) /* Mangled header, drop. */
659 if (icmph
->type
== ICMP_DEST_UNREACH
) {
660 switch (icmph
->code
& 15) {
661 case ICMP_NET_UNREACH
:
662 case ICMP_HOST_UNREACH
:
663 case ICMP_PROT_UNREACH
:
664 case ICMP_PORT_UNREACH
:
666 case ICMP_FRAG_NEEDED
:
667 if (ipv4_config
.no_pmtu_disc
) {
668 LIMIT_NETDEBUG(KERN_INFO
"ICMP: %pI4: fragmentation needed and DF set.\n",
671 info
= ip_rt_frag_needed(net
, iph
,
672 ntohs(icmph
->un
.frag
.mtu
),
679 LIMIT_NETDEBUG(KERN_INFO
"ICMP: %pI4: Source Route Failed.\n",
685 if (icmph
->code
> NR_ICMP_UNREACH
)
687 } else if (icmph
->type
== ICMP_PARAMETERPROB
)
688 info
= ntohl(icmph
->un
.gateway
) >> 24;
691 * Throw it at our lower layers
693 * RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
695 * RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
697 * RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
702 * Check the other end isnt violating RFC 1122. Some routers send
703 * bogus responses to broadcast frames. If you see this message
704 * first check your netmask matches at both ends, if it does then
705 * get the other vendor to fix their kit.
708 if (!net
->ipv4
.sysctl_icmp_ignore_bogus_error_responses
&&
709 inet_addr_type(net
, iph
->daddr
) == RTN_BROADCAST
) {
711 printk(KERN_WARNING
"%pI4 sent an invalid ICMP "
713 "error to a broadcast: %pI4 on %s\n",
715 icmph
->type
, icmph
->code
,
721 /* Checkin full IP header plus 8 bytes of protocol to
722 * avoid additional coding at protocol handlers.
724 if (!pskb_may_pull(skb
, iph
->ihl
* 4 + 8))
727 iph
= (struct iphdr
*)skb
->data
;
728 protocol
= iph
->protocol
;
731 * Deliver ICMP message to raw sockets. Pretty useless feature?
733 raw_icmp_error(skb
, protocol
, info
);
735 hash
= protocol
& (MAX_INET_PROTOS
- 1);
737 ipprot
= rcu_dereference(inet_protos
[hash
]);
738 if (ipprot
&& ipprot
->err_handler
)
739 ipprot
->err_handler(skb
, info
);
745 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
751 * Handle ICMP_REDIRECT.
754 static void icmp_redirect(struct sk_buff
*skb
)
758 if (skb
->len
< sizeof(struct iphdr
))
762 * Get the copied header of the packet that caused the redirect
764 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
767 iph
= (struct iphdr
*)skb
->data
;
769 switch (icmp_hdr(skb
)->code
& 7) {
771 case ICMP_REDIR_NETTOS
:
773 * As per RFC recommendations now handle it as a host redirect.
775 case ICMP_REDIR_HOST
:
776 case ICMP_REDIR_HOSTTOS
:
777 ip_rt_redirect(ip_hdr(skb
)->saddr
, iph
->daddr
,
778 icmp_hdr(skb
)->un
.gateway
,
779 iph
->saddr
, skb
->dev
);
785 ICMP_INC_STATS_BH(dev_net(skb
->dev
), ICMP_MIB_INERRORS
);
790 * Handle ICMP_ECHO ("ping") requests.
792 * RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
794 * RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
795 * included in the reply.
796 * RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
797 * echo requests, MUST have default=NOT.
798 * See also WRT handling of options once they are done and working.
801 static void icmp_echo(struct sk_buff
*skb
)
805 net
= dev_net(skb_dst(skb
)->dev
);
806 if (!net
->ipv4
.sysctl_icmp_echo_ignore_all
) {
807 struct icmp_bxm icmp_param
;
809 icmp_param
.data
.icmph
= *icmp_hdr(skb
);
810 icmp_param
.data
.icmph
.type
= ICMP_ECHOREPLY
;
811 icmp_param
.skb
= skb
;
812 icmp_param
.offset
= 0;
813 icmp_param
.data_len
= skb
->len
;
814 icmp_param
.head_len
= sizeof(struct icmphdr
);
815 icmp_reply(&icmp_param
, skb
);
820 * Handle ICMP Timestamp requests.
821 * RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
822 * SHOULD be in the kernel for minimum random latency.
823 * MUST be accurate to a few minutes.
824 * MUST be updated at least at 15Hz.
826 static void icmp_timestamp(struct sk_buff
*skb
)
829 struct icmp_bxm icmp_param
;
837 * Fill in the current time as ms since midnight UT:
840 icmp_param
.data
.times
[1] = htonl((tv
.tv_sec
% 86400) * MSEC_PER_SEC
+
841 tv
.tv_nsec
/ NSEC_PER_MSEC
);
842 icmp_param
.data
.times
[2] = icmp_param
.data
.times
[1];
843 if (skb_copy_bits(skb
, 0, &icmp_param
.data
.times
[0], 4))
845 icmp_param
.data
.icmph
= *icmp_hdr(skb
);
846 icmp_param
.data
.icmph
.type
= ICMP_TIMESTAMPREPLY
;
847 icmp_param
.data
.icmph
.code
= 0;
848 icmp_param
.skb
= skb
;
849 icmp_param
.offset
= 0;
850 icmp_param
.data_len
= 0;
851 icmp_param
.head_len
= sizeof(struct icmphdr
) + 12;
852 icmp_reply(&icmp_param
, skb
);
856 ICMP_INC_STATS_BH(dev_net(skb_dst(skb
)->dev
), ICMP_MIB_INERRORS
);
862 * Handle ICMP_ADDRESS_MASK requests. (RFC950)
864 * RFC1122 (3.2.2.9). A host MUST only send replies to
865 * ADDRESS_MASK requests if it's been configured as an address mask
866 * agent. Receiving a request doesn't constitute implicit permission to
867 * act as one. Of course, implementing this correctly requires (SHOULD)
868 * a way to turn the functionality on and off. Another one for sysctl(),
871 * RFC1812 (4.3.3.9). A router MUST implement it.
872 * A router SHOULD have switch turning it on/off.
873 * This switch MUST be ON by default.
875 * Gratuitous replies, zero-source replies are not implemented,
876 * that complies with RFC. DO NOT implement them!!! All the idea
877 * of broadcast addrmask replies as specified in RFC950 is broken.
878 * The problem is that it is not uncommon to have several prefixes
879 * on one physical interface. Moreover, addrmask agent can even be
880 * not aware of existing another prefixes.
881 * If source is zero, addrmask agent cannot choose correct prefix.
882 * Gratuitous mask announcements suffer from the same problem.
883 * RFC1812 explains it, but still allows to use ADDRMASK,
884 * that is pretty silly. --ANK
886 * All these rules are so bizarre, that I removed kernel addrmask
887 * support at all. It is wrong, it is obsolete, nobody uses it in
890 * Furthermore you can do it with a usermode address agent program
894 static void icmp_address(struct sk_buff
*skb
)
898 printk(KERN_DEBUG
"a guy asks for address mask. Who is it?\n");
903 * RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
904 * loudly if an inconsistency is found.
905 * called with rcu_read_lock()
908 static void icmp_address_reply(struct sk_buff
*skb
)
910 struct rtable
*rt
= skb_rtable(skb
);
911 struct net_device
*dev
= skb
->dev
;
912 struct in_device
*in_dev
;
913 struct in_ifaddr
*ifa
;
915 if (skb
->len
< 4 || !(rt
->rt_flags
&RTCF_DIRECTSRC
))
918 in_dev
= __in_dev_get_rcu(dev
);
922 if (in_dev
->ifa_list
&&
923 IN_DEV_LOG_MARTIANS(in_dev
) &&
924 IN_DEV_FORWARD(in_dev
)) {
927 mp
= skb_header_pointer(skb
, 0, sizeof(_mask
), &_mask
);
929 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
) {
930 if (*mp
== ifa
->ifa_mask
&&
931 inet_ifa_match(rt
->rt_src
, ifa
))
934 if (!ifa
&& net_ratelimit()) {
935 printk(KERN_INFO
"Wrong address mask %pI4 from %s/%pI4\n",
936 mp
, dev
->name
, &rt
->rt_src
);
941 static void icmp_discard(struct sk_buff
*skb
)
946 * Deal with incoming ICMP packets.
948 int icmp_rcv(struct sk_buff
*skb
)
950 struct icmphdr
*icmph
;
951 struct rtable
*rt
= skb_rtable(skb
);
952 struct net
*net
= dev_net(rt
->dst
.dev
);
954 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
955 struct sec_path
*sp
= skb_sec_path(skb
);
958 if (!(sp
&& sp
->xvec
[sp
->len
- 1]->props
.flags
&
962 if (!pskb_may_pull(skb
, sizeof(*icmph
) + sizeof(struct iphdr
)))
965 nh
= skb_network_offset(skb
);
966 skb_set_network_header(skb
, sizeof(*icmph
));
968 if (!xfrm4_policy_check_reverse(NULL
, XFRM_POLICY_IN
, skb
))
971 skb_set_network_header(skb
, nh
);
974 ICMP_INC_STATS_BH(net
, ICMP_MIB_INMSGS
);
976 switch (skb
->ip_summed
) {
977 case CHECKSUM_COMPLETE
:
978 if (!csum_fold(skb
->csum
))
983 if (__skb_checksum_complete(skb
))
987 if (!pskb_pull(skb
, sizeof(*icmph
)))
990 icmph
= icmp_hdr(skb
);
992 ICMPMSGIN_INC_STATS_BH(net
, icmph
->type
);
994 * 18 is the highest 'known' ICMP type. Anything else is a mystery
996 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
999 if (icmph
->type
> NR_ICMP_TYPES
)
1004 * Parse the ICMP message
1007 if (rt
->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
1009 * RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
1010 * silently ignored (we let user decide with a sysctl).
1011 * RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1012 * discarded if to broadcast/multicast.
1014 if ((icmph
->type
== ICMP_ECHO
||
1015 icmph
->type
== ICMP_TIMESTAMP
) &&
1016 net
->ipv4
.sysctl_icmp_echo_ignore_broadcasts
) {
1019 if (icmph
->type
!= ICMP_ECHO
&&
1020 icmph
->type
!= ICMP_TIMESTAMP
&&
1021 icmph
->type
!= ICMP_ADDRESS
&&
1022 icmph
->type
!= ICMP_ADDRESSREPLY
) {
1027 icmp_pointers
[icmph
->type
].handler(skb
);
1033 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
1038 * This table is the definition of how we handle ICMP.
1040 static const struct icmp_control icmp_pointers
[NR_ICMP_TYPES
+ 1] = {
1041 [ICMP_ECHOREPLY
] = {
1042 .handler
= icmp_discard
,
1045 .handler
= icmp_discard
,
1049 .handler
= icmp_discard
,
1052 [ICMP_DEST_UNREACH
] = {
1053 .handler
= icmp_unreach
,
1056 [ICMP_SOURCE_QUENCH
] = {
1057 .handler
= icmp_unreach
,
1061 .handler
= icmp_redirect
,
1065 .handler
= icmp_discard
,
1069 .handler
= icmp_discard
,
1073 .handler
= icmp_echo
,
1076 .handler
= icmp_discard
,
1080 .handler
= icmp_discard
,
1083 [ICMP_TIME_EXCEEDED
] = {
1084 .handler
= icmp_unreach
,
1087 [ICMP_PARAMETERPROB
] = {
1088 .handler
= icmp_unreach
,
1091 [ICMP_TIMESTAMP
] = {
1092 .handler
= icmp_timestamp
,
1094 [ICMP_TIMESTAMPREPLY
] = {
1095 .handler
= icmp_discard
,
1097 [ICMP_INFO_REQUEST
] = {
1098 .handler
= icmp_discard
,
1100 [ICMP_INFO_REPLY
] = {
1101 .handler
= icmp_discard
,
1104 .handler
= icmp_address
,
1106 [ICMP_ADDRESSREPLY
] = {
1107 .handler
= icmp_address_reply
,
1111 static void __net_exit
icmp_sk_exit(struct net
*net
)
1115 for_each_possible_cpu(i
)
1116 inet_ctl_sock_destroy(net
->ipv4
.icmp_sk
[i
]);
1117 kfree(net
->ipv4
.icmp_sk
);
1118 net
->ipv4
.icmp_sk
= NULL
;
1121 static int __net_init
icmp_sk_init(struct net
*net
)
1126 kzalloc(nr_cpu_ids
* sizeof(struct sock
*), GFP_KERNEL
);
1127 if (net
->ipv4
.icmp_sk
== NULL
)
1130 for_each_possible_cpu(i
) {
1133 err
= inet_ctl_sock_create(&sk
, PF_INET
,
1134 SOCK_RAW
, IPPROTO_ICMP
, net
);
1138 net
->ipv4
.icmp_sk
[i
] = sk
;
1140 /* Enough space for 2 64K ICMP packets, including
1141 * sk_buff struct overhead.
1144 (2 * ((64 * 1024) + sizeof(struct sk_buff
)));
1147 * Speedup sock_wfree()
1149 sock_set_flag(sk
, SOCK_USE_WRITE_QUEUE
);
1150 inet_sk(sk
)->pmtudisc
= IP_PMTUDISC_DONT
;
1153 /* Control parameters for ECHO replies. */
1154 net
->ipv4
.sysctl_icmp_echo_ignore_all
= 0;
1155 net
->ipv4
.sysctl_icmp_echo_ignore_broadcasts
= 1;
1157 /* Control parameter - ignore bogus broadcast responses? */
1158 net
->ipv4
.sysctl_icmp_ignore_bogus_error_responses
= 1;
1161 * Configurable global rate limit.
1163 * ratelimit defines tokens/packet consumed for dst->rate_token
1164 * bucket ratemask defines which icmp types are ratelimited by
1165 * setting it's bit position.
1168 * dest unreachable (3), source quench (4),
1169 * time exceeded (11), parameter problem (12)
1172 net
->ipv4
.sysctl_icmp_ratelimit
= 1 * HZ
;
1173 net
->ipv4
.sysctl_icmp_ratemask
= 0x1818;
1174 net
->ipv4
.sysctl_icmp_errors_use_inbound_ifaddr
= 0;
1179 for_each_possible_cpu(i
)
1180 inet_ctl_sock_destroy(net
->ipv4
.icmp_sk
[i
]);
1181 kfree(net
->ipv4
.icmp_sk
);
1185 static struct pernet_operations __net_initdata icmp_sk_ops
= {
1186 .init
= icmp_sk_init
,
1187 .exit
= icmp_sk_exit
,
1190 int __init
icmp_init(void)
1192 return register_pernet_subsys(&icmp_sk_ops
);