1 /* (C) 1999-2001 Paul `Rusty' Russell
2 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
3 * (C) 2006-2010 Patrick McHardy <kaber@trash.net>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
10 #include <linux/types.h>
11 #include <linux/timer.h>
12 #include <linux/netfilter.h>
14 #include <linux/icmp.h>
15 #include <linux/seq_file.h>
17 #include <net/checksum.h>
18 #include <linux/netfilter_ipv4.h>
19 #include <net/netfilter/nf_conntrack_tuple.h>
20 #include <net/netfilter/nf_conntrack_l4proto.h>
21 #include <net/netfilter/nf_conntrack_core.h>
22 #include <net/netfilter/nf_conntrack_zones.h>
23 #include <net/netfilter/nf_log.h>
25 static unsigned int nf_ct_icmp_timeout __read_mostly
= 30*HZ
;
27 static inline struct nf_icmp_net
*icmp_pernet(struct net
*net
)
29 return &net
->ct
.nf_ct_proto
.icmp
;
32 static bool icmp_pkt_to_tuple(const struct sk_buff
*skb
, unsigned int dataoff
,
33 struct net
*net
, struct nf_conntrack_tuple
*tuple
)
35 const struct icmphdr
*hp
;
38 hp
= skb_header_pointer(skb
, dataoff
, sizeof(_hdr
), &_hdr
);
42 tuple
->dst
.u
.icmp
.type
= hp
->type
;
43 tuple
->src
.u
.icmp
.id
= hp
->un
.echo
.id
;
44 tuple
->dst
.u
.icmp
.code
= hp
->code
;
49 /* Add 1; spaces filled with 0. */
50 static const u_int8_t invmap
[] = {
51 [ICMP_ECHO
] = ICMP_ECHOREPLY
+ 1,
52 [ICMP_ECHOREPLY
] = ICMP_ECHO
+ 1,
53 [ICMP_TIMESTAMP
] = ICMP_TIMESTAMPREPLY
+ 1,
54 [ICMP_TIMESTAMPREPLY
] = ICMP_TIMESTAMP
+ 1,
55 [ICMP_INFO_REQUEST
] = ICMP_INFO_REPLY
+ 1,
56 [ICMP_INFO_REPLY
] = ICMP_INFO_REQUEST
+ 1,
57 [ICMP_ADDRESS
] = ICMP_ADDRESSREPLY
+ 1,
58 [ICMP_ADDRESSREPLY
] = ICMP_ADDRESS
+ 1
61 static bool icmp_invert_tuple(struct nf_conntrack_tuple
*tuple
,
62 const struct nf_conntrack_tuple
*orig
)
64 if (orig
->dst
.u
.icmp
.type
>= sizeof(invmap
) ||
65 !invmap
[orig
->dst
.u
.icmp
.type
])
68 tuple
->src
.u
.icmp
.id
= orig
->src
.u
.icmp
.id
;
69 tuple
->dst
.u
.icmp
.type
= invmap
[orig
->dst
.u
.icmp
.type
] - 1;
70 tuple
->dst
.u
.icmp
.code
= orig
->dst
.u
.icmp
.code
;
74 static unsigned int *icmp_get_timeouts(struct net
*net
)
76 return &icmp_pernet(net
)->timeout
;
79 /* Returns verdict for packet, or -1 for invalid. */
80 static int icmp_packet(struct nf_conn
*ct
,
81 const struct sk_buff
*skb
,
83 enum ip_conntrack_info ctinfo
,
85 unsigned int *timeout
)
87 /* Do not immediately delete the connection after the first
88 successful reply to avoid excessive conntrackd traffic
89 and also to handle correctly ICMP echo reply duplicates. */
90 nf_ct_refresh_acct(ct
, ctinfo
, skb
, *timeout
);
95 /* Called when a new connection for this protocol found. */
96 static bool icmp_new(struct nf_conn
*ct
, const struct sk_buff
*skb
,
97 unsigned int dataoff
, unsigned int *timeouts
)
99 static const u_int8_t valid_new
[] = {
101 [ICMP_TIMESTAMP
] = 1,
102 [ICMP_INFO_REQUEST
] = 1,
106 if (ct
->tuplehash
[0].tuple
.dst
.u
.icmp
.type
>= sizeof(valid_new
) ||
107 !valid_new
[ct
->tuplehash
[0].tuple
.dst
.u
.icmp
.type
]) {
108 /* Can't create a new ICMP `conn' with this. */
109 pr_debug("icmp: can't create new conn with type %u\n",
110 ct
->tuplehash
[0].tuple
.dst
.u
.icmp
.type
);
111 nf_ct_dump_tuple_ip(&ct
->tuplehash
[0].tuple
);
117 /* Returns conntrack if it dealt with ICMP, and filled in skb fields */
119 icmp_error_message(struct net
*net
, struct nf_conn
*tmpl
, struct sk_buff
*skb
,
120 unsigned int hooknum
)
122 struct nf_conntrack_tuple innertuple
, origtuple
;
123 const struct nf_conntrack_l4proto
*innerproto
;
124 const struct nf_conntrack_tuple_hash
*h
;
125 const struct nf_conntrack_zone
*zone
;
126 enum ip_conntrack_info ctinfo
;
127 struct nf_conntrack_zone tmp
;
129 WARN_ON(skb_nfct(skb
));
130 zone
= nf_ct_zone_tmpl(tmpl
, skb
, &tmp
);
132 /* Are they talking about one of our connections? */
133 if (!nf_ct_get_tuplepr(skb
,
134 skb_network_offset(skb
) + ip_hdrlen(skb
)
135 + sizeof(struct icmphdr
),
136 PF_INET
, net
, &origtuple
)) {
137 pr_debug("icmp_error_message: failed to get tuple\n");
141 /* rcu_read_lock()ed by nf_hook_thresh */
142 innerproto
= __nf_ct_l4proto_find(PF_INET
, origtuple
.dst
.protonum
);
144 /* Ordinarily, we'd expect the inverted tupleproto, but it's
145 been preserved inside the ICMP. */
146 if (!nf_ct_invert_tuple(&innertuple
, &origtuple
,
147 &nf_conntrack_l3proto_ipv4
, innerproto
)) {
148 pr_debug("icmp_error_message: no match\n");
152 ctinfo
= IP_CT_RELATED
;
154 h
= nf_conntrack_find_get(net
, zone
, &innertuple
);
156 pr_debug("icmp_error_message: no match\n");
160 if (NF_CT_DIRECTION(h
) == IP_CT_DIR_REPLY
)
161 ctinfo
+= IP_CT_IS_REPLY
;
163 /* Update skb to refer to this connection */
164 nf_ct_set(skb
, nf_ct_tuplehash_to_ctrack(h
), ctinfo
);
168 /* Small and modified version of icmp_rcv */
170 icmp_error(struct net
*net
, struct nf_conn
*tmpl
,
171 struct sk_buff
*skb
, unsigned int dataoff
,
172 u8 pf
, unsigned int hooknum
)
174 const struct icmphdr
*icmph
;
177 /* Not enough header? */
178 icmph
= skb_header_pointer(skb
, ip_hdrlen(skb
), sizeof(_ih
), &_ih
);
180 if (LOG_INVALID(net
, IPPROTO_ICMP
))
181 nf_log_packet(net
, PF_INET
, 0, skb
, NULL
, NULL
,
182 NULL
, "nf_ct_icmp: short packet ");
186 /* See ip_conntrack_proto_tcp.c */
187 if (net
->ct
.sysctl_checksum
&& hooknum
== NF_INET_PRE_ROUTING
&&
188 nf_ip_checksum(skb
, hooknum
, dataoff
, 0)) {
189 if (LOG_INVALID(net
, IPPROTO_ICMP
))
190 nf_log_packet(net
, PF_INET
, 0, skb
, NULL
, NULL
, NULL
,
191 "nf_ct_icmp: bad HW ICMP checksum ");
196 * 18 is the highest 'known' ICMP type. Anything else is a mystery
198 * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently
201 if (icmph
->type
> NR_ICMP_TYPES
) {
202 if (LOG_INVALID(net
, IPPROTO_ICMP
))
203 nf_log_packet(net
, PF_INET
, 0, skb
, NULL
, NULL
, NULL
,
204 "nf_ct_icmp: invalid ICMP type ");
208 /* Need to track icmp error message? */
209 if (icmph
->type
!= ICMP_DEST_UNREACH
&&
210 icmph
->type
!= ICMP_SOURCE_QUENCH
&&
211 icmph
->type
!= ICMP_TIME_EXCEEDED
&&
212 icmph
->type
!= ICMP_PARAMETERPROB
&&
213 icmph
->type
!= ICMP_REDIRECT
)
216 return icmp_error_message(net
, tmpl
, skb
, hooknum
);
219 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
221 #include <linux/netfilter/nfnetlink.h>
222 #include <linux/netfilter/nfnetlink_conntrack.h>
224 static int icmp_tuple_to_nlattr(struct sk_buff
*skb
,
225 const struct nf_conntrack_tuple
*t
)
227 if (nla_put_be16(skb
, CTA_PROTO_ICMP_ID
, t
->src
.u
.icmp
.id
) ||
228 nla_put_u8(skb
, CTA_PROTO_ICMP_TYPE
, t
->dst
.u
.icmp
.type
) ||
229 nla_put_u8(skb
, CTA_PROTO_ICMP_CODE
, t
->dst
.u
.icmp
.code
))
230 goto nla_put_failure
;
237 static const struct nla_policy icmp_nla_policy
[CTA_PROTO_MAX
+1] = {
238 [CTA_PROTO_ICMP_TYPE
] = { .type
= NLA_U8
},
239 [CTA_PROTO_ICMP_CODE
] = { .type
= NLA_U8
},
240 [CTA_PROTO_ICMP_ID
] = { .type
= NLA_U16
},
243 static int icmp_nlattr_to_tuple(struct nlattr
*tb
[],
244 struct nf_conntrack_tuple
*tuple
)
246 if (!tb
[CTA_PROTO_ICMP_TYPE
] ||
247 !tb
[CTA_PROTO_ICMP_CODE
] ||
248 !tb
[CTA_PROTO_ICMP_ID
])
251 tuple
->dst
.u
.icmp
.type
= nla_get_u8(tb
[CTA_PROTO_ICMP_TYPE
]);
252 tuple
->dst
.u
.icmp
.code
= nla_get_u8(tb
[CTA_PROTO_ICMP_CODE
]);
253 tuple
->src
.u
.icmp
.id
= nla_get_be16(tb
[CTA_PROTO_ICMP_ID
]);
255 if (tuple
->dst
.u
.icmp
.type
>= sizeof(invmap
) ||
256 !invmap
[tuple
->dst
.u
.icmp
.type
])
262 static int icmp_nlattr_tuple_size(void)
264 return nla_policy_len(icmp_nla_policy
, CTA_PROTO_MAX
+ 1);
268 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
270 #include <linux/netfilter/nfnetlink.h>
271 #include <linux/netfilter/nfnetlink_cttimeout.h>
273 static int icmp_timeout_nlattr_to_obj(struct nlattr
*tb
[],
274 struct net
*net
, void *data
)
276 unsigned int *timeout
= data
;
277 struct nf_icmp_net
*in
= icmp_pernet(net
);
279 if (tb
[CTA_TIMEOUT_ICMP_TIMEOUT
]) {
281 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_ICMP_TIMEOUT
])) * HZ
;
283 /* Set default ICMP timeout. */
284 *timeout
= in
->timeout
;
290 icmp_timeout_obj_to_nlattr(struct sk_buff
*skb
, const void *data
)
292 const unsigned int *timeout
= data
;
294 if (nla_put_be32(skb
, CTA_TIMEOUT_ICMP_TIMEOUT
, htonl(*timeout
/ HZ
)))
295 goto nla_put_failure
;
302 static const struct nla_policy
303 icmp_timeout_nla_policy
[CTA_TIMEOUT_ICMP_MAX
+1] = {
304 [CTA_TIMEOUT_ICMP_TIMEOUT
] = { .type
= NLA_U32
},
306 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
309 static struct ctl_table icmp_sysctl_table
[] = {
311 .procname
= "nf_conntrack_icmp_timeout",
312 .maxlen
= sizeof(unsigned int),
314 .proc_handler
= proc_dointvec_jiffies
,
318 #endif /* CONFIG_SYSCTL */
320 static int icmp_kmemdup_sysctl_table(struct nf_proto_net
*pn
,
321 struct nf_icmp_net
*in
)
324 pn
->ctl_table
= kmemdup(icmp_sysctl_table
,
325 sizeof(icmp_sysctl_table
),
330 pn
->ctl_table
[0].data
= &in
->timeout
;
335 static int icmp_init_net(struct net
*net
, u_int16_t proto
)
337 struct nf_icmp_net
*in
= icmp_pernet(net
);
338 struct nf_proto_net
*pn
= &in
->pn
;
340 in
->timeout
= nf_ct_icmp_timeout
;
342 return icmp_kmemdup_sysctl_table(pn
, in
);
345 static struct nf_proto_net
*icmp_get_net_proto(struct net
*net
)
347 return &net
->ct
.nf_ct_proto
.icmp
.pn
;
350 struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp __read_mostly
=
353 .l4proto
= IPPROTO_ICMP
,
354 .pkt_to_tuple
= icmp_pkt_to_tuple
,
355 .invert_tuple
= icmp_invert_tuple
,
356 .packet
= icmp_packet
,
357 .get_timeouts
= icmp_get_timeouts
,
362 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
363 .tuple_to_nlattr
= icmp_tuple_to_nlattr
,
364 .nlattr_tuple_size
= icmp_nlattr_tuple_size
,
365 .nlattr_to_tuple
= icmp_nlattr_to_tuple
,
366 .nla_policy
= icmp_nla_policy
,
368 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
370 .nlattr_to_obj
= icmp_timeout_nlattr_to_obj
,
371 .obj_to_nlattr
= icmp_timeout_obj_to_nlattr
,
372 .nlattr_max
= CTA_TIMEOUT_ICMP_MAX
,
373 .obj_size
= sizeof(unsigned int),
374 .nla_policy
= icmp_timeout_nla_policy
,
376 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
377 .init_net
= icmp_init_net
,
378 .get_net_proto
= icmp_get_net_proto
,