net/ipv4/netfilter/nf_conntrack_proto_icmp.c

   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
  22 static unsigned long nf_ct_icmp_timeout __read_mostly = 30*HZ;
  23
  24 static int icmp_pkt_to_tuple(const struct sk_buff *skb,
  25                              unsigned int dataoff,
  26                              struct nf_conntrack_tuple *tuple)
  27 {
  28         struct icmphdr _hdr, *hp;
  29
  30         hp = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
  31         if (hp == NULL)
  32                 return 0;
  33
  34         tuple->dst.u.icmp.type = hp->type;
  35         tuple->src.u.icmp.id = hp->un.echo.id;
  36         tuple->dst.u.icmp.code = hp->code;
  37
  38         return 1;
  39 }
  40
  41 /* Add 1; spaces filled with 0. */
  42 static const u_int8_t invmap[] = {
  43         [ICMP_ECHO] = ICMP_ECHOREPLY + 1,
  44         [ICMP_ECHOREPLY] = ICMP_ECHO + 1,
  45         [ICMP_TIMESTAMP] = ICMP_TIMESTAMPREPLY + 1,
  46         [ICMP_TIMESTAMPREPLY] = ICMP_TIMESTAMP + 1,
  47         [ICMP_INFO_REQUEST] = ICMP_INFO_REPLY + 1,
  48         [ICMP_INFO_REPLY] = ICMP_INFO_REQUEST + 1,
  49         [ICMP_ADDRESS] = ICMP_ADDRESSREPLY + 1,
  50         [ICMP_ADDRESSREPLY] = ICMP_ADDRESS + 1
  51 };
  52
  53 static int icmp_invert_tuple(struct nf_conntrack_tuple *tuple,
  54                              const struct nf_conntrack_tuple *orig)
  55 {
  56         if (orig->dst.u.icmp.type >= sizeof(invmap)
  57             || !invmap[orig->dst.u.icmp.type])
  58                 return 0;
  59
  60         tuple->src.u.icmp.id = orig->src.u.icmp.id;
  61         tuple->dst.u.icmp.type = invmap[orig->dst.u.icmp.type] - 1;
  62         tuple->dst.u.icmp.code = orig->dst.u.icmp.code;
  63         return 1;
  64 }
  65
  66 /* Print out the per-protocol part of the tuple. */
  67 static int icmp_print_tuple(struct seq_file *s,
  68                             const struct nf_conntrack_tuple *tuple)
  69 {
  70         return seq_printf(s, "type=%u code=%u id=%u ",
  71                           tuple->dst.u.icmp.type,
  72                           tuple->dst.u.icmp.code,
  73                           ntohs(tuple->src.u.icmp.id));
  74 }
  75
  76 /* Print out the private part of the conntrack. */
  77 static int icmp_print_conntrack(struct seq_file *s,
  78                                 const struct nf_conn *conntrack)
  79 {
  80         return 0;
  81 }
  82
  83 /* Returns verdict for packet, or -1 for invalid. */
  84 static int icmp_packet(struct nf_conn *ct,
  85                        const struct sk_buff *skb,
  86                        unsigned int dataoff,
  87                        enum ip_conntrack_info ctinfo,
  88                        int pf,
  89                        unsigned int hooknum)
  90 {
  91         /* Try to delete connection immediately after all replies:
  92            won't actually vanish as we still have skb, and del_timer
  93            means this will only run once even if count hits zero twice
  94            (theoretically possible with SMP) */
  95         if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) {
  96                 if (atomic_dec_and_test(&ct->proto.icmp.count)
  97                     && del_timer(&ct->timeout))
  98                         ct->timeout.function((unsigned long)ct);
  99         } else {
 100                 atomic_inc(&ct->proto.icmp.count);
 101                 nf_conntrack_event_cache(IPCT_PROTOINFO_VOLATILE, skb);
 102                 nf_ct_refresh_acct(ct, ctinfo, skb, nf_ct_icmp_timeout);
 103         }
 104
 105         return NF_ACCEPT;
 106 }
 107
 108 /* Called when a new connection for this protocol found. */
 109 static int icmp_new(struct nf_conn *conntrack,
 110                     const struct sk_buff *skb, unsigned int dataoff)
 111 {
 112         static const u_int8_t valid_new[] = {
 113                 [ICMP_ECHO] = 1,
 114                 [ICMP_TIMESTAMP] = 1,
 115                 [ICMP_INFO_REQUEST] = 1,
 116                 [ICMP_ADDRESS] = 1
 117         };
 118
 119         if (conntrack->tuplehash[0].tuple.dst.u.icmp.type >= sizeof(valid_new)
 120             || !valid_new[conntrack->tuplehash[0].tuple.dst.u.icmp.type]) {
 121                 /* Can't create a new ICMP `conn' with this. */
 122                 pr_debug("icmp: can't create new conn with type %u\n",
 123                          conntrack->tuplehash[0].tuple.dst.u.icmp.type);
 124                 NF_CT_DUMP_TUPLE(&conntrack->tuplehash[0].tuple);
 125                 return 0;
 126         }
 127         atomic_set(&conntrack->proto.icmp.count, 0);
 128         return 1;
 129 }
 130
 131 extern struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4;
 132 /* Returns conntrack if it dealt with ICMP, and filled in skb fields */
 133 static int
 134 icmp_error_message(struct sk_buff *skb,
 135                  enum ip_conntrack_info *ctinfo,
 136                  unsigned int hooknum)
 137 {
 138         struct nf_conntrack_tuple innertuple, origtuple;
 139         struct nf_conntrack_l4proto *innerproto;
 140         struct nf_conntrack_tuple_hash *h;
 141
 142         NF_CT_ASSERT(skb->nfct == NULL);
 143
 144         /* Are they talking about one of our connections? */
 145         if (!nf_ct_get_tuplepr(skb,
 146                                skb_network_offset(skb) + ip_hdrlen(skb)
 147                                                        + sizeof(struct icmphdr),
 148                                PF_INET, &origtuple)) {
 149                 pr_debug("icmp_error_message: failed to get tuple\n");
 150                 return -NF_ACCEPT;
 151         }
 152
 153         /* rcu_read_lock()ed by nf_hook_slow */
 154         innerproto = __nf_ct_l4proto_find(PF_INET, origtuple.dst.protonum);
 155
 156         /* Ordinarily, we'd expect the inverted tupleproto, but it's
 157            been preserved inside the ICMP. */
 158         if (!nf_ct_invert_tuple(&innertuple, &origtuple,
 159                                 &nf_conntrack_l3proto_ipv4, innerproto)) {
 160                 pr_debug("icmp_error_message: no match\n");
 161                 return -NF_ACCEPT;
 162         }
 163
 164         *ctinfo = IP_CT_RELATED;
 165
 166         h = nf_conntrack_find_get(&innertuple);
 167         if (!h) {
 168                 pr_debug("icmp_error_message: no match\n");
 169                 return -NF_ACCEPT;
 170         }
 171
 172         if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
 173                 *ctinfo += IP_CT_IS_REPLY;
 174
 175         /* Update skb to refer to this connection */
 176         skb->nfct = &nf_ct_tuplehash_to_ctrack(h)->ct_general;
 177         skb->nfctinfo = *ctinfo;
 178         return -NF_ACCEPT;
 179 }
 180
 181 /* Small and modified version of icmp_rcv */
 182 static int
 183 icmp_error(struct sk_buff *skb, unsigned int dataoff,
 184            enum ip_conntrack_info *ctinfo, int pf, unsigned int hooknum)
 185 {
 186         struct icmphdr _ih, *icmph;
 187
 188         /* Not enough header? */
 189         icmph = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_ih), &_ih);
 190         if (icmph == NULL) {
 191                 if (LOG_INVALID(IPPROTO_ICMP))
 192                         nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL,
 193                                       "nf_ct_icmp: short packet ");
 194                 return -NF_ACCEPT;
 195         }
 196
 197         /* See ip_conntrack_proto_tcp.c */
 198         if (nf_conntrack_checksum && hooknum == NF_IP_PRE_ROUTING &&
 199             nf_ip_checksum(skb, hooknum, dataoff, 0)) {
 200                 if (LOG_INVALID(IPPROTO_ICMP))
 201                         nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL,
 202                                       "nf_ct_icmp: bad HW ICMP checksum ");
 203                 return -NF_ACCEPT;
 204         }
 205
 206         /*
 207          *      18 is the highest 'known' ICMP type. Anything else is a mystery
 208          *
 209          *      RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently
 210          *                discarded.
 211          */
 212         if (icmph->type > NR_ICMP_TYPES) {
 213                 if (LOG_INVALID(IPPROTO_ICMP))
 214                         nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL,
 215                                       "nf_ct_icmp: invalid ICMP type ");
 216                 return -NF_ACCEPT;
 217         }
 218
 219         /* Need to track icmp error message? */
 220         if (icmph->type != ICMP_DEST_UNREACH
 221             && icmph->type != ICMP_SOURCE_QUENCH
 222             && icmph->type != ICMP_TIME_EXCEEDED
 223             && icmph->type != ICMP_PARAMETERPROB
 224             && icmph->type != ICMP_REDIRECT)
 225                 return NF_ACCEPT;
 226
 227         return icmp_error_message(skb, ctinfo, hooknum);
 228 }
 229
 230 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 231
 232 #include <linux/netfilter/nfnetlink.h>
 233 #include <linux/netfilter/nfnetlink_conntrack.h>
 234
 235 static int icmp_tuple_to_nfattr(struct sk_buff *skb,
 236                                 const struct nf_conntrack_tuple *t)
 237 {
 238         NLA_PUT(skb, CTA_PROTO_ICMP_ID, sizeof(u_int16_t),
 239                 &t->src.u.icmp.id);
 240         NLA_PUT(skb, CTA_PROTO_ICMP_TYPE, sizeof(u_int8_t),
 241                 &t->dst.u.icmp.type);
 242         NLA_PUT(skb, CTA_PROTO_ICMP_CODE, sizeof(u_int8_t),
 243                 &t->dst.u.icmp.code);
 244
 245         return 0;
 246
 247 nla_put_failure:
 248         return -1;
 249 }
 250
 251 static const size_t cta_min_proto[CTA_PROTO_MAX+1] = {
 252         [CTA_PROTO_ICMP_TYPE]   = sizeof(u_int8_t),
 253         [CTA_PROTO_ICMP_CODE]   = sizeof(u_int8_t),
 254         [CTA_PROTO_ICMP_ID]     = sizeof(u_int16_t)
 255 };
 256
 257 static int icmp_nfattr_to_tuple(struct nlattr *tb[],
 258                                 struct nf_conntrack_tuple *tuple)
 259 {
 260         if (!tb[CTA_PROTO_ICMP_TYPE]
 261             || !tb[CTA_PROTO_ICMP_CODE]
 262             || !tb[CTA_PROTO_ICMP_ID])
 263                 return -EINVAL;
 264
 265         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 266                 return -EINVAL;
 267
 268         tuple->dst.u.icmp.type =
 269                         *(u_int8_t *)nla_data(tb[CTA_PROTO_ICMP_TYPE]);
 270         tuple->dst.u.icmp.code =
 271                         *(u_int8_t *)nla_data(tb[CTA_PROTO_ICMP_CODE]);
 272         tuple->src.u.icmp.id =
 273                         *(__be16 *)nla_data(tb[CTA_PROTO_ICMP_ID]);
 274
 275         if (tuple->dst.u.icmp.type >= sizeof(invmap)
 276             || !invmap[tuple->dst.u.icmp.type])
 277                 return -EINVAL;
 278
 279         return 0;
 280 }
 281 #endif
 282
 283 #ifdef CONFIG_SYSCTL
 284 static struct ctl_table_header *icmp_sysctl_header;
 285 static struct ctl_table icmp_sysctl_table[] = {
 286         {
 287                 .ctl_name       = NET_NF_CONNTRACK_ICMP_TIMEOUT,
 288                 .procname       = "nf_conntrack_icmp_timeout",
 289                 .data           = &nf_ct_icmp_timeout,
 290                 .maxlen         = sizeof(unsigned int),
 291                 .mode           = 0644,
 292                 .proc_handler   = &proc_dointvec_jiffies,
 293         },
 294         {
 295                 .ctl_name = 0
 296         }
 297 };
 298 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
 299 static struct ctl_table icmp_compat_sysctl_table[] = {
 300         {
 301                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 302                 .procname       = "ip_conntrack_icmp_timeout",
 303                 .data           = &nf_ct_icmp_timeout,
 304                 .maxlen         = sizeof(unsigned int),
 305                 .mode           = 0644,
 306                 .proc_handler   = &proc_dointvec_jiffies,
 307         },
 308         {
 309                 .ctl_name = 0
 310         }
 311 };
 312 #endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
 313 #endif /* CONFIG_SYSCTL */
 314
 315 struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp __read_mostly =
 316 {
 317         .l3proto                = PF_INET,
 318         .l4proto                = IPPROTO_ICMP,
 319         .name                   = "icmp",
 320         .pkt_to_tuple           = icmp_pkt_to_tuple,
 321         .invert_tuple           = icmp_invert_tuple,
 322         .print_tuple            = icmp_print_tuple,
 323         .print_conntrack        = icmp_print_conntrack,
 324         .packet                 = icmp_packet,
 325         .new                    = icmp_new,
 326         .error                  = icmp_error,
 327         .destroy                = NULL,
 328         .me                     = NULL,
 329 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 330         .tuple_to_nfattr        = icmp_tuple_to_nfattr,
 331         .nfattr_to_tuple        = icmp_nfattr_to_tuple,
 332 #endif
 333 #ifdef CONFIG_SYSCTL
 334         .ctl_table_header       = &icmp_sysctl_header,
 335         .ctl_table              = icmp_sysctl_table,
 336 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
 337         .ctl_compat_table       = icmp_compat_sysctl_table,
 338 #endif
 339 #endif
 340 };