| 1 | /* (C) 1999-2001 Paul `Rusty' Russell |
| 2 | * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org> |
| 3 | * |
| 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. |
| 7 | */ |
| 8 | |
| 9 | #include <linux/types.h> |
| 10 | #include <linux/timer.h> |
| 11 | #include <linux/netfilter.h> |
| 12 | #include <linux/in.h> |
| 13 | #include <linux/icmp.h> |
| 14 | #include <linux/seq_file.h> |
| 15 | #include <net/ip.h> |
| 16 | #include <net/checksum.h> |
| 17 | #include <linux/netfilter_ipv4.h> |
| 18 | #include <net/netfilter/nf_conntrack_tuple.h> |
| 19 | #include <net/netfilter/nf_conntrack_l4proto.h> |
| 20 | #include <net/netfilter/nf_conntrack_core.h> |
| 21 | #include <net/netfilter/nf_conntrack_zones.h> |
| 22 | #include <net/netfilter/nf_log.h> |
| 23 | |
| 24 | static unsigned int nf_ct_icmp_timeout __read_mostly = 30*HZ; |
| 25 | |
| 26 | static bool icmp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff, |
| 27 | struct nf_conntrack_tuple *tuple) |
| 28 | { |
| 29 | const struct icmphdr *hp; |
| 30 | struct icmphdr _hdr; |
| 31 | |
| 32 | hp = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr); |
| 33 | if (hp == NULL) |
| 34 | return false; |
| 35 | |
| 36 | tuple->dst.u.icmp.type = hp->type; |
| 37 | tuple->src.u.icmp.id = hp->un.echo.id; |
| 38 | tuple->dst.u.icmp.code = hp->code; |
| 39 | |
| 40 | return true; |
| 41 | } |
| 42 | |
| 43 | /* Add 1; spaces filled with 0. */ |
| 44 | static const u_int8_t invmap[] = { |
| 45 | [ICMP_ECHO] = ICMP_ECHOREPLY + 1, |
| 46 | [ICMP_ECHOREPLY] = ICMP_ECHO + 1, |
| 47 | [ICMP_TIMESTAMP] = ICMP_TIMESTAMPREPLY + 1, |
| 48 | [ICMP_TIMESTAMPREPLY] = ICMP_TIMESTAMP + 1, |
| 49 | [ICMP_INFO_REQUEST] = ICMP_INFO_REPLY + 1, |
| 50 | [ICMP_INFO_REPLY] = ICMP_INFO_REQUEST + 1, |
| 51 | [ICMP_ADDRESS] = ICMP_ADDRESSREPLY + 1, |
| 52 | [ICMP_ADDRESSREPLY] = ICMP_ADDRESS + 1 |
| 53 | }; |
| 54 | |
| 55 | static bool icmp_invert_tuple(struct nf_conntrack_tuple *tuple, |
| 56 | const struct nf_conntrack_tuple *orig) |
| 57 | { |
| 58 | if (orig->dst.u.icmp.type >= sizeof(invmap) || |
| 59 | !invmap[orig->dst.u.icmp.type]) |
| 60 | return false; |
| 61 | |
| 62 | tuple->src.u.icmp.id = orig->src.u.icmp.id; |
| 63 | tuple->dst.u.icmp.type = invmap[orig->dst.u.icmp.type] - 1; |
| 64 | tuple->dst.u.icmp.code = orig->dst.u.icmp.code; |
| 65 | return true; |
| 66 | } |
| 67 | |
| 68 | /* Print out the per-protocol part of the tuple. */ |
| 69 | static int icmp_print_tuple(struct seq_file *s, |
| 70 | const struct nf_conntrack_tuple *tuple) |
| 71 | { |
| 72 | return seq_printf(s, "type=%u code=%u id=%u ", |
| 73 | tuple->dst.u.icmp.type, |
| 74 | tuple->dst.u.icmp.code, |
| 75 | ntohs(tuple->src.u.icmp.id)); |
| 76 | } |
| 77 | |
| 78 | /* Returns verdict for packet, or -1 for invalid. */ |
| 79 | static int icmp_packet(struct nf_conn *ct, |
| 80 | const struct sk_buff *skb, |
| 81 | unsigned int dataoff, |
| 82 | enum ip_conntrack_info ctinfo, |
| 83 | u_int8_t pf, |
| 84 | unsigned int hooknum) |
| 85 | { |
| 86 | /* Do not immediately delete the connection after the first |
| 87 | successful reply to avoid excessive conntrackd traffic |
| 88 | and also to handle correctly ICMP echo reply duplicates. */ |
| 89 | nf_ct_refresh_acct(ct, ctinfo, skb, nf_ct_icmp_timeout); |
| 90 | |
| 91 | return NF_ACCEPT; |
| 92 | } |
| 93 | |
| 94 | /* Called when a new connection for this protocol found. */ |
| 95 | static bool icmp_new(struct nf_conn *ct, const struct sk_buff *skb, |
| 96 | unsigned int dataoff) |
| 97 | { |
| 98 | static const u_int8_t valid_new[] = { |
| 99 | [ICMP_ECHO] = 1, |
| 100 | [ICMP_TIMESTAMP] = 1, |
| 101 | [ICMP_INFO_REQUEST] = 1, |
| 102 | [ICMP_ADDRESS] = 1 |
| 103 | }; |
| 104 | |
| 105 | if (ct->tuplehash[0].tuple.dst.u.icmp.type >= sizeof(valid_new) || |
| 106 | !valid_new[ct->tuplehash[0].tuple.dst.u.icmp.type]) { |
| 107 | /* Can't create a new ICMP `conn' with this. */ |
| 108 | pr_debug("icmp: can't create new conn with type %u\n", |
| 109 | ct->tuplehash[0].tuple.dst.u.icmp.type); |
| 110 | nf_ct_dump_tuple_ip(&ct->tuplehash[0].tuple); |
| 111 | return false; |
| 112 | } |
| 113 | return true; |
| 114 | } |
| 115 | |
| 116 | /* Returns conntrack if it dealt with ICMP, and filled in skb fields */ |
| 117 | static int |
| 118 | icmp_error_message(struct net *net, struct nf_conn *tmpl, struct sk_buff *skb, |
| 119 | enum ip_conntrack_info *ctinfo, |
| 120 | unsigned int hooknum) |
| 121 | { |
| 122 | struct nf_conntrack_tuple innertuple, origtuple; |
| 123 | const struct nf_conntrack_l4proto *innerproto; |
| 124 | const struct nf_conntrack_tuple_hash *h; |
| 125 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
| 126 | |
| 127 | NF_CT_ASSERT(skb->nfct == NULL); |
| 128 | |
| 129 | /* Are they talking about one of our connections? */ |
| 130 | if (!nf_ct_get_tuplepr(skb, |
| 131 | skb_network_offset(skb) + ip_hdrlen(skb) |
| 132 | + sizeof(struct icmphdr), |
| 133 | PF_INET, &origtuple)) { |
| 134 | pr_debug("icmp_error_message: failed to get tuple\n"); |
| 135 | return -NF_ACCEPT; |
| 136 | } |
| 137 | |
| 138 | /* rcu_read_lock()ed by nf_hook_slow */ |
| 139 | innerproto = __nf_ct_l4proto_find(PF_INET, origtuple.dst.protonum); |
| 140 | |
| 141 | /* Ordinarily, we'd expect the inverted tupleproto, but it's |
| 142 | been preserved inside the ICMP. */ |
| 143 | if (!nf_ct_invert_tuple(&innertuple, &origtuple, |
| 144 | &nf_conntrack_l3proto_ipv4, innerproto)) { |
| 145 | pr_debug("icmp_error_message: no match\n"); |
| 146 | return -NF_ACCEPT; |
| 147 | } |
| 148 | |
| 149 | *ctinfo = IP_CT_RELATED; |
| 150 | |
| 151 | h = nf_conntrack_find_get(net, zone, &innertuple); |
| 152 | if (!h) { |
| 153 | pr_debug("icmp_error_message: no match\n"); |
| 154 | return -NF_ACCEPT; |
| 155 | } |
| 156 | |
| 157 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) |
| 158 | *ctinfo += IP_CT_IS_REPLY; |
| 159 | |
| 160 | /* Update skb to refer to this connection */ |
| 161 | skb->nfct = &nf_ct_tuplehash_to_ctrack(h)->ct_general; |
| 162 | skb->nfctinfo = *ctinfo; |
| 163 | return -NF_ACCEPT; |
| 164 | } |
| 165 | |
| 166 | /* Small and modified version of icmp_rcv */ |
| 167 | static int |
| 168 | icmp_error(struct net *net, struct nf_conn *tmpl, |
| 169 | struct sk_buff *skb, unsigned int dataoff, |
| 170 | enum ip_conntrack_info *ctinfo, u_int8_t pf, unsigned int hooknum) |
| 171 | { |
| 172 | const struct icmphdr *icmph; |
| 173 | struct icmphdr _ih; |
| 174 | |
| 175 | /* Not enough header? */ |
| 176 | icmph = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_ih), &_ih); |
| 177 | if (icmph == NULL) { |
| 178 | if (LOG_INVALID(net, IPPROTO_ICMP)) |
| 179 | nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL, |
| 180 | "nf_ct_icmp: short packet "); |
| 181 | return -NF_ACCEPT; |
| 182 | } |
| 183 | |
| 184 | /* See ip_conntrack_proto_tcp.c */ |
| 185 | if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING && |
| 186 | nf_ip_checksum(skb, hooknum, dataoff, 0)) { |
| 187 | if (LOG_INVALID(net, IPPROTO_ICMP)) |
| 188 | nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL, |
| 189 | "nf_ct_icmp: bad HW ICMP checksum "); |
| 190 | return -NF_ACCEPT; |
| 191 | } |
| 192 | |
| 193 | /* |
| 194 | * 18 is the highest 'known' ICMP type. Anything else is a mystery |
| 195 | * |
| 196 | * RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently |
| 197 | * discarded. |
| 198 | */ |
| 199 | if (icmph->type > NR_ICMP_TYPES) { |
| 200 | if (LOG_INVALID(net, IPPROTO_ICMP)) |
| 201 | nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL, |
| 202 | "nf_ct_icmp: invalid ICMP type "); |
| 203 | return -NF_ACCEPT; |
| 204 | } |
| 205 | |
| 206 | /* Need to track icmp error message? */ |
| 207 | if (icmph->type != ICMP_DEST_UNREACH && |
| 208 | icmph->type != ICMP_SOURCE_QUENCH && |
| 209 | icmph->type != ICMP_TIME_EXCEEDED && |
| 210 | icmph->type != ICMP_PARAMETERPROB && |
| 211 | icmph->type != ICMP_REDIRECT) |
| 212 | return NF_ACCEPT; |
| 213 | |
| 214 | return icmp_error_message(net, tmpl, skb, ctinfo, hooknum); |
| 215 | } |
| 216 | |
| 217 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
| 218 | |
| 219 | #include <linux/netfilter/nfnetlink.h> |
| 220 | #include <linux/netfilter/nfnetlink_conntrack.h> |
| 221 | |
| 222 | static int icmp_tuple_to_nlattr(struct sk_buff *skb, |
| 223 | const struct nf_conntrack_tuple *t) |
| 224 | { |
| 225 | NLA_PUT_BE16(skb, CTA_PROTO_ICMP_ID, t->src.u.icmp.id); |
| 226 | NLA_PUT_U8(skb, CTA_PROTO_ICMP_TYPE, t->dst.u.icmp.type); |
| 227 | NLA_PUT_U8(skb, CTA_PROTO_ICMP_CODE, t->dst.u.icmp.code); |
| 228 | |
| 229 | return 0; |
| 230 | |
| 231 | nla_put_failure: |
| 232 | return -1; |
| 233 | } |
| 234 | |
| 235 | static const struct nla_policy icmp_nla_policy[CTA_PROTO_MAX+1] = { |
| 236 | [CTA_PROTO_ICMP_TYPE] = { .type = NLA_U8 }, |
| 237 | [CTA_PROTO_ICMP_CODE] = { .type = NLA_U8 }, |
| 238 | [CTA_PROTO_ICMP_ID] = { .type = NLA_U16 }, |
| 239 | }; |
| 240 | |
| 241 | static int icmp_nlattr_to_tuple(struct nlattr *tb[], |
| 242 | struct nf_conntrack_tuple *tuple) |
| 243 | { |
| 244 | if (!tb[CTA_PROTO_ICMP_TYPE] || |
| 245 | !tb[CTA_PROTO_ICMP_CODE] || |
| 246 | !tb[CTA_PROTO_ICMP_ID]) |
| 247 | return -EINVAL; |
| 248 | |
| 249 | tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMP_TYPE]); |
| 250 | tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMP_CODE]); |
| 251 | tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMP_ID]); |
| 252 | |
| 253 | if (tuple->dst.u.icmp.type >= sizeof(invmap) || |
| 254 | !invmap[tuple->dst.u.icmp.type]) |
| 255 | return -EINVAL; |
| 256 | |
| 257 | return 0; |
| 258 | } |
| 259 | |
| 260 | static int icmp_nlattr_tuple_size(void) |
| 261 | { |
| 262 | return nla_policy_len(icmp_nla_policy, CTA_PROTO_MAX + 1); |
| 263 | } |
| 264 | #endif |
| 265 | |
| 266 | #ifdef CONFIG_SYSCTL |
| 267 | static struct ctl_table_header *icmp_sysctl_header; |
| 268 | static struct ctl_table icmp_sysctl_table[] = { |
| 269 | { |
| 270 | .procname = "nf_conntrack_icmp_timeout", |
| 271 | .data = &nf_ct_icmp_timeout, |
| 272 | .maxlen = sizeof(unsigned int), |
| 273 | .mode = 0644, |
| 274 | .proc_handler = proc_dointvec_jiffies, |
| 275 | }, |
| 276 | { } |
| 277 | }; |
| 278 | #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT |
| 279 | static struct ctl_table icmp_compat_sysctl_table[] = { |
| 280 | { |
| 281 | .procname = "ip_conntrack_icmp_timeout", |
| 282 | .data = &nf_ct_icmp_timeout, |
| 283 | .maxlen = sizeof(unsigned int), |
| 284 | .mode = 0644, |
| 285 | .proc_handler = proc_dointvec_jiffies, |
| 286 | }, |
| 287 | { } |
| 288 | }; |
| 289 | #endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */ |
| 290 | #endif /* CONFIG_SYSCTL */ |
| 291 | |
| 292 | struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp __read_mostly = |
| 293 | { |
| 294 | .l3proto = PF_INET, |
| 295 | .l4proto = IPPROTO_ICMP, |
| 296 | .name = "icmp", |
| 297 | .pkt_to_tuple = icmp_pkt_to_tuple, |
| 298 | .invert_tuple = icmp_invert_tuple, |
| 299 | .print_tuple = icmp_print_tuple, |
| 300 | .packet = icmp_packet, |
| 301 | .new = icmp_new, |
| 302 | .error = icmp_error, |
| 303 | .destroy = NULL, |
| 304 | .me = NULL, |
| 305 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
| 306 | .tuple_to_nlattr = icmp_tuple_to_nlattr, |
| 307 | .nlattr_tuple_size = icmp_nlattr_tuple_size, |
| 308 | .nlattr_to_tuple = icmp_nlattr_to_tuple, |
| 309 | .nla_policy = icmp_nla_policy, |
| 310 | #endif |
| 311 | #ifdef CONFIG_SYSCTL |
| 312 | .ctl_table_header = &icmp_sysctl_header, |
| 313 | .ctl_table = icmp_sysctl_table, |
| 314 | #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT |
| 315 | .ctl_compat_table = icmp_compat_sysctl_table, |
| 316 | #endif |
| 317 | #endif |
| 318 | }; |