2 * Copyright (C)2003,2004 USAGI/WIDE Project
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
9 * Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
12 #include <linux/types.h>
13 #include <linux/timer.h>
14 #include <linux/module.h>
15 #include <linux/netfilter.h>
16 #include <linux/in6.h>
17 #include <linux/icmpv6.h>
18 #include <linux/ipv6.h>
20 #include <net/ip6_checksum.h>
21 #include <linux/seq_file.h>
22 #include <linux/netfilter_ipv6.h>
23 #include <net/netfilter/nf_conntrack_tuple.h>
24 #include <net/netfilter/nf_conntrack_l4proto.h>
25 #include <net/netfilter/nf_conntrack_core.h>
26 #include <net/netfilter/nf_conntrack_timeout.h>
27 #include <net/netfilter/nf_conntrack_zones.h>
28 #include <net/netfilter/ipv6/nf_conntrack_icmpv6.h>
29 #include <net/netfilter/nf_log.h>
31 static const unsigned int nf_ct_icmpv6_timeout = 30*HZ;
33 static bool icmpv6_pkt_to_tuple(const struct sk_buff *skb,
36 struct nf_conntrack_tuple *tuple)
38 const struct icmp6hdr *hp;
41 hp = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
44 tuple->dst.u.icmp.type = hp->icmp6_type;
45 tuple->src.u.icmp.id = hp->icmp6_identifier;
46 tuple->dst.u.icmp.code = hp->icmp6_code;
51 /* Add 1; spaces filled with 0. */
52 static const u_int8_t invmap[] = {
53 [ICMPV6_ECHO_REQUEST - 128] = ICMPV6_ECHO_REPLY + 1,
54 [ICMPV6_ECHO_REPLY - 128] = ICMPV6_ECHO_REQUEST + 1,
55 [ICMPV6_NI_QUERY - 128] = ICMPV6_NI_REPLY + 1,
56 [ICMPV6_NI_REPLY - 128] = ICMPV6_NI_QUERY + 1
59 static const u_int8_t noct_valid_new[] = {
60 [ICMPV6_MGM_QUERY - 130] = 1,
61 [ICMPV6_MGM_REPORT - 130] = 1,
62 [ICMPV6_MGM_REDUCTION - 130] = 1,
63 [NDISC_ROUTER_SOLICITATION - 130] = 1,
64 [NDISC_ROUTER_ADVERTISEMENT - 130] = 1,
65 [NDISC_NEIGHBOUR_SOLICITATION - 130] = 1,
66 [NDISC_NEIGHBOUR_ADVERTISEMENT - 130] = 1,
67 [ICMPV6_MLD2_REPORT - 130] = 1
70 static bool icmpv6_invert_tuple(struct nf_conntrack_tuple *tuple,
71 const struct nf_conntrack_tuple *orig)
73 int type = orig->dst.u.icmp.type - 128;
74 if (type < 0 || type >= sizeof(invmap) || !invmap[type])
77 tuple->src.u.icmp.id = orig->src.u.icmp.id;
78 tuple->dst.u.icmp.type = invmap[type] - 1;
79 tuple->dst.u.icmp.code = orig->dst.u.icmp.code;
83 static unsigned int *icmpv6_get_timeouts(struct net *net)
85 return &nf_icmpv6_pernet(net)->timeout;
88 /* Returns verdict for packet, or -1 for invalid. */
89 static int icmpv6_packet(struct nf_conn *ct,
92 enum ip_conntrack_info ctinfo,
93 const struct nf_hook_state *state)
95 unsigned int *timeout = nf_ct_timeout_lookup(ct);
96 static const u8 valid_new[] = {
97 [ICMPV6_ECHO_REQUEST - 128] = 1,
98 [ICMPV6_NI_QUERY - 128] = 1
101 if (state->pf != NFPROTO_IPV6)
104 if (!nf_ct_is_confirmed(ct)) {
105 int type = ct->tuplehash[0].tuple.dst.u.icmp.type - 128;
107 if (type < 0 || type >= sizeof(valid_new) || !valid_new[type]) {
108 /* Can't create a new ICMPv6 `conn' with this. */
109 pr_debug("icmpv6: can't create new conn with type %u\n",
111 nf_ct_dump_tuple_ipv6(&ct->tuplehash[0].tuple);
117 timeout = icmpv6_get_timeouts(nf_ct_net(ct));
119 /* Do not immediately delete the connection after the first
120 successful reply to avoid excessive conntrackd traffic
121 and also to handle correctly ICMP echo reply duplicates. */
122 nf_ct_refresh_acct(ct, ctinfo, skb, *timeout);
128 icmpv6_error_message(struct net *net, struct nf_conn *tmpl,
130 unsigned int icmp6off)
132 struct nf_conntrack_tuple intuple, origtuple;
133 const struct nf_conntrack_tuple_hash *h;
134 const struct nf_conntrack_l4proto *inproto;
135 enum ip_conntrack_info ctinfo;
136 struct nf_conntrack_zone tmp;
138 WARN_ON(skb_nfct(skb));
140 /* Are they talking about one of our connections? */
141 if (!nf_ct_get_tuplepr(skb,
142 skb_network_offset(skb)
143 + sizeof(struct ipv6hdr)
144 + sizeof(struct icmp6hdr),
145 PF_INET6, net, &origtuple)) {
146 pr_debug("icmpv6_error: Can't get tuple\n");
150 /* rcu_read_lock()ed by nf_hook_thresh */
151 inproto = __nf_ct_l4proto_find(origtuple.dst.protonum);
153 /* Ordinarily, we'd expect the inverted tupleproto, but it's
154 been preserved inside the ICMP. */
155 if (!nf_ct_invert_tuple(&intuple, &origtuple, inproto)) {
156 pr_debug("icmpv6_error: Can't invert tuple\n");
160 ctinfo = IP_CT_RELATED;
162 h = nf_conntrack_find_get(net, nf_ct_zone_tmpl(tmpl, skb, &tmp),
165 pr_debug("icmpv6_error: no match\n");
168 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
169 ctinfo += IP_CT_IS_REPLY;
172 /* Update skb to refer to this connection */
173 nf_ct_set(skb, nf_ct_tuplehash_to_ctrack(h), ctinfo);
177 static void icmpv6_error_log(const struct sk_buff *skb,
178 const struct nf_hook_state *state,
181 nf_l4proto_log_invalid(skb, state->net, state->pf,
182 IPPROTO_ICMPV6, "%s", msg);
185 int nf_conntrack_icmpv6_error(struct nf_conn *tmpl,
187 unsigned int dataoff,
188 const struct nf_hook_state *state)
190 const struct icmp6hdr *icmp6h;
194 icmp6h = skb_header_pointer(skb, dataoff, sizeof(_ih), &_ih);
195 if (icmp6h == NULL) {
196 icmpv6_error_log(skb, state, "short packet");
200 if (state->hook == NF_INET_PRE_ROUTING &&
201 state->net->ct.sysctl_checksum &&
202 nf_ip6_checksum(skb, state->hook, dataoff, IPPROTO_ICMPV6)) {
203 icmpv6_error_log(skb, state, "ICMPv6 checksum failed");
207 type = icmp6h->icmp6_type - 130;
208 if (type >= 0 && type < sizeof(noct_valid_new) &&
209 noct_valid_new[type]) {
210 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
214 /* is not error message ? */
215 if (icmp6h->icmp6_type >= 128)
218 return icmpv6_error_message(state->net, tmpl, skb, dataoff);
221 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
223 #include <linux/netfilter/nfnetlink.h>
224 #include <linux/netfilter/nfnetlink_conntrack.h>
225 static int icmpv6_tuple_to_nlattr(struct sk_buff *skb,
226 const struct nf_conntrack_tuple *t)
228 if (nla_put_be16(skb, CTA_PROTO_ICMPV6_ID, t->src.u.icmp.id) ||
229 nla_put_u8(skb, CTA_PROTO_ICMPV6_TYPE, t->dst.u.icmp.type) ||
230 nla_put_u8(skb, CTA_PROTO_ICMPV6_CODE, t->dst.u.icmp.code))
231 goto nla_put_failure;
238 static const struct nla_policy icmpv6_nla_policy[CTA_PROTO_MAX+1] = {
239 [CTA_PROTO_ICMPV6_TYPE] = { .type = NLA_U8 },
240 [CTA_PROTO_ICMPV6_CODE] = { .type = NLA_U8 },
241 [CTA_PROTO_ICMPV6_ID] = { .type = NLA_U16 },
244 static int icmpv6_nlattr_to_tuple(struct nlattr *tb[],
245 struct nf_conntrack_tuple *tuple)
247 if (!tb[CTA_PROTO_ICMPV6_TYPE] ||
248 !tb[CTA_PROTO_ICMPV6_CODE] ||
249 !tb[CTA_PROTO_ICMPV6_ID])
252 tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMPV6_TYPE]);
253 tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMPV6_CODE]);
254 tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMPV6_ID]);
256 if (tuple->dst.u.icmp.type < 128 ||
257 tuple->dst.u.icmp.type - 128 >= sizeof(invmap) ||
258 !invmap[tuple->dst.u.icmp.type - 128])
264 static unsigned int icmpv6_nlattr_tuple_size(void)
266 static unsigned int size __read_mostly;
269 size = nla_policy_len(icmpv6_nla_policy, CTA_PROTO_MAX + 1);
275 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
277 #include <linux/netfilter/nfnetlink.h>
278 #include <linux/netfilter/nfnetlink_cttimeout.h>
280 static int icmpv6_timeout_nlattr_to_obj(struct nlattr *tb[],
281 struct net *net, void *data)
283 unsigned int *timeout = data;
284 struct nf_icmp_net *in = nf_icmpv6_pernet(net);
287 timeout = icmpv6_get_timeouts(net);
288 if (tb[CTA_TIMEOUT_ICMPV6_TIMEOUT]) {
290 ntohl(nla_get_be32(tb[CTA_TIMEOUT_ICMPV6_TIMEOUT])) * HZ;
292 /* Set default ICMPv6 timeout. */
293 *timeout = in->timeout;
299 icmpv6_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
301 const unsigned int *timeout = data;
303 if (nla_put_be32(skb, CTA_TIMEOUT_ICMPV6_TIMEOUT, htonl(*timeout / HZ)))
304 goto nla_put_failure;
311 static const struct nla_policy
312 icmpv6_timeout_nla_policy[CTA_TIMEOUT_ICMPV6_MAX+1] = {
313 [CTA_TIMEOUT_ICMPV6_TIMEOUT] = { .type = NLA_U32 },
315 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
318 static struct ctl_table icmpv6_sysctl_table[] = {
320 .procname = "nf_conntrack_icmpv6_timeout",
321 .maxlen = sizeof(unsigned int),
323 .proc_handler = proc_dointvec_jiffies,
327 #endif /* CONFIG_SYSCTL */
329 static int icmpv6_kmemdup_sysctl_table(struct nf_proto_net *pn,
330 struct nf_icmp_net *in)
333 pn->ctl_table = kmemdup(icmpv6_sysctl_table,
334 sizeof(icmpv6_sysctl_table),
339 pn->ctl_table[0].data = &in->timeout;
344 static int icmpv6_init_net(struct net *net)
346 struct nf_icmp_net *in = nf_icmpv6_pernet(net);
347 struct nf_proto_net *pn = &in->pn;
349 in->timeout = nf_ct_icmpv6_timeout;
351 return icmpv6_kmemdup_sysctl_table(pn, in);
354 static struct nf_proto_net *icmpv6_get_net_proto(struct net *net)
356 return &net->ct.nf_ct_proto.icmpv6.pn;
359 const struct nf_conntrack_l4proto nf_conntrack_l4proto_icmpv6 =
361 .l4proto = IPPROTO_ICMPV6,
362 .pkt_to_tuple = icmpv6_pkt_to_tuple,
363 .invert_tuple = icmpv6_invert_tuple,
364 .packet = icmpv6_packet,
365 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
366 .tuple_to_nlattr = icmpv6_tuple_to_nlattr,
367 .nlattr_tuple_size = icmpv6_nlattr_tuple_size,
368 .nlattr_to_tuple = icmpv6_nlattr_to_tuple,
369 .nla_policy = icmpv6_nla_policy,
371 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
373 .nlattr_to_obj = icmpv6_timeout_nlattr_to_obj,
374 .obj_to_nlattr = icmpv6_timeout_obj_to_nlattr,
375 .nlattr_max = CTA_TIMEOUT_ICMP_MAX,
376 .obj_size = sizeof(unsigned int),
377 .nla_policy = icmpv6_timeout_nla_policy,
379 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
380 .init_net = icmpv6_init_net,
381 .get_net_proto = icmpv6_get_net_proto,