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  * (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   4  *
   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.
   8  */
   9
  10 #include <linux/types.h>
  11 #include <linux/timer.h>
  12 #include <linux/netfilter.h>
  13 #include <linux/in.h>
  14 #include <linux/icmp.h>
  15 #include <linux/seq_file.h>
  16 #include <net/ip.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>
  24
  25 static unsigned int nf_ct_icmp_timeout __read_mostly = 30*HZ;
  26
  27 static inline struct nf_icmp_net *icmp_pernet(struct net *net)
  28 {
  29         return &net->ct.nf_ct_proto.icmp;
  30 }
  31
  32 static bool icmp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
  33                               struct net *net, struct nf_conntrack_tuple *tuple)
  34 {
  35         const struct icmphdr *hp;
  36         struct icmphdr _hdr;
  37
  38         hp = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
  39         if (hp == NULL)
  40                 return false;
  41
  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;
  45
  46         return true;
  47 }
  48
  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
  59 };
  60
  61 static bool icmp_invert_tuple(struct nf_conntrack_tuple *tuple,
  62                               const struct nf_conntrack_tuple *orig)
  63 {
  64         if (orig->dst.u.icmp.type >= sizeof(invmap) ||
  65             !invmap[orig->dst.u.icmp.type])
  66                 return false;
  67
  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;
  71         return true;
  72 }
  73
  74 static unsigned int *icmp_get_timeouts(struct net *net)
  75 {
  76         return &icmp_pernet(net)->timeout;
  77 }
  78
  79 /* Returns verdict for packet, or -1 for invalid. */
  80 static int icmp_packet(struct nf_conn *ct,
  81                        const struct sk_buff *skb,
  82                        unsigned int dataoff,
  83                        enum ip_conntrack_info ctinfo,
  84                        u_int8_t pf,
  85                        unsigned int *timeout)
  86 {
  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);
  91
  92         return NF_ACCEPT;
  93 }
  94
  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)
  98 {
  99         static const u_int8_t valid_new[] = {
 100                 [ICMP_ECHO] = 1,
 101                 [ICMP_TIMESTAMP] = 1,
 102                 [ICMP_INFO_REQUEST] = 1,
 103                 [ICMP_ADDRESS] = 1
 104         };
 105
 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);
 112                 return false;
 113         }
 114         return true;
 115 }
 116
 117 /* Returns conntrack if it dealt with ICMP, and filled in skb fields */
 118 static int
 119 icmp_error_message(struct net *net, struct nf_conn *tmpl, struct sk_buff *skb,
 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         const struct nf_conntrack_zone *zone;
 126         enum ip_conntrack_info ctinfo;
 127         struct nf_conntrack_zone tmp;
 128
 129         WARN_ON(skb_nfct(skb));
 130         zone = nf_ct_zone_tmpl(tmpl, skb, &tmp);
 131
 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");
 138                 return -NF_ACCEPT;
 139         }
 140
 141         /* rcu_read_lock()ed by nf_hook_thresh */
 142         innerproto = __nf_ct_l4proto_find(PF_INET, origtuple.dst.protonum);
 143
 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");
 149                 return -NF_ACCEPT;
 150         }
 151
 152         ctinfo = IP_CT_RELATED;
 153
 154         h = nf_conntrack_find_get(net, zone, &innertuple);
 155         if (!h) {
 156                 pr_debug("icmp_error_message: no match\n");
 157                 return -NF_ACCEPT;
 158         }
 159
 160         if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
 161                 ctinfo += IP_CT_IS_REPLY;
 162
 163         /* Update skb to refer to this connection */
 164         nf_ct_set(skb, nf_ct_tuplehash_to_ctrack(h), ctinfo);
 165         return NF_ACCEPT;
 166 }
 167
 168 static void icmp_error_log(const struct sk_buff *skb, struct net *net,
 169                            u8 pf, const char *msg)
 170 {
 171         nf_l4proto_log_invalid(skb, net, pf, IPPROTO_ICMP, "%s", msg);
 172 }
 173
 174 /* Small and modified version of icmp_rcv */
 175 static int
 176 icmp_error(struct net *net, struct nf_conn *tmpl,
 177            struct sk_buff *skb, unsigned int dataoff,
 178            u8 pf, unsigned int hooknum)
 179 {
 180         const struct icmphdr *icmph;
 181         struct icmphdr _ih;
 182
 183         /* Not enough header? */
 184         icmph = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_ih), &_ih);
 185         if (icmph == NULL) {
 186                 icmp_error_log(skb, net, pf, "short packet");
 187                 return -NF_ACCEPT;
 188         }
 189
 190         /* See ip_conntrack_proto_tcp.c */
 191         if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
 192             nf_ip_checksum(skb, hooknum, dataoff, 0)) {
 193                 icmp_error_log(skb, net, pf, "bad hw icmp checksum");
 194                 return -NF_ACCEPT;
 195         }
 196
 197         /*
 198          *      18 is the highest 'known' ICMP type. Anything else is a mystery
 199          *
 200          *      RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently
 201          *                discarded.
 202          */
 203         if (icmph->type > NR_ICMP_TYPES) {
 204                 icmp_error_log(skb, net, pf, "invalid icmp type");
 205                 return -NF_ACCEPT;
 206         }
 207
 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)
 214                 return NF_ACCEPT;
 215
 216         return icmp_error_message(net, tmpl, skb, hooknum);
 217 }
 218
 219 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 220
 221 #include <linux/netfilter/nfnetlink.h>
 222 #include <linux/netfilter/nfnetlink_conntrack.h>
 223
 224 static int icmp_tuple_to_nlattr(struct sk_buff *skb,
 225                                 const struct nf_conntrack_tuple *t)
 226 {
 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;
 231         return 0;
 232
 233 nla_put_failure:
 234         return -1;
 235 }
 236
 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 },
 241 };
 242
 243 static int icmp_nlattr_to_tuple(struct nlattr *tb[],
 244                                 struct nf_conntrack_tuple *tuple)
 245 {
 246         if (!tb[CTA_PROTO_ICMP_TYPE] ||
 247             !tb[CTA_PROTO_ICMP_CODE] ||
 248             !tb[CTA_PROTO_ICMP_ID])
 249                 return -EINVAL;
 250
 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]);
 254
 255         if (tuple->dst.u.icmp.type >= sizeof(invmap) ||
 256             !invmap[tuple->dst.u.icmp.type])
 257                 return -EINVAL;
 258
 259         return 0;
 260 }
 261
 262 static int icmp_nlattr_tuple_size(void)
 263 {
 264         return nla_policy_len(icmp_nla_policy, CTA_PROTO_MAX + 1);
 265 }
 266 #endif
 267
 268 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 269
 270 #include <linux/netfilter/nfnetlink.h>
 271 #include <linux/netfilter/nfnetlink_cttimeout.h>
 272
 273 static int icmp_timeout_nlattr_to_obj(struct nlattr *tb[],
 274                                       struct net *net, void *data)
 275 {
 276         unsigned int *timeout = data;
 277         struct nf_icmp_net *in = icmp_pernet(net);
 278
 279         if (tb[CTA_TIMEOUT_ICMP_TIMEOUT]) {
 280                 *timeout =
 281                         ntohl(nla_get_be32(tb[CTA_TIMEOUT_ICMP_TIMEOUT])) * HZ;
 282         } else {
 283                 /* Set default ICMP timeout. */
 284                 *timeout = in->timeout;
 285         }
 286         return 0;
 287 }
 288
 289 static int
 290 icmp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
 291 {
 292         const unsigned int *timeout = data;
 293
 294         if (nla_put_be32(skb, CTA_TIMEOUT_ICMP_TIMEOUT, htonl(*timeout / HZ)))
 295                 goto nla_put_failure;
 296         return 0;
 297
 298 nla_put_failure:
 299         return -ENOSPC;
 300 }
 301
 302 static const struct nla_policy
 303 icmp_timeout_nla_policy[CTA_TIMEOUT_ICMP_MAX+1] = {
 304         [CTA_TIMEOUT_ICMP_TIMEOUT]      = { .type = NLA_U32 },
 305 };
 306 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
 307
 308 #ifdef CONFIG_SYSCTL
 309 static struct ctl_table icmp_sysctl_table[] = {
 310         {
 311                 .procname       = "nf_conntrack_icmp_timeout",
 312                 .maxlen         = sizeof(unsigned int),
 313                 .mode           = 0644,
 314                 .proc_handler   = proc_dointvec_jiffies,
 315         },
 316         { }
 317 };
 318 #endif /* CONFIG_SYSCTL */
 319
 320 static int icmp_kmemdup_sysctl_table(struct nf_proto_net *pn,
 321                                      struct nf_icmp_net *in)
 322 {
 323 #ifdef CONFIG_SYSCTL
 324         pn->ctl_table = kmemdup(icmp_sysctl_table,
 325                                 sizeof(icmp_sysctl_table),
 326                                 GFP_KERNEL);
 327         if (!pn->ctl_table)
 328                 return -ENOMEM;
 329
 330         pn->ctl_table[0].data = &in->timeout;
 331 #endif
 332         return 0;
 333 }
 334
 335 static int icmp_init_net(struct net *net, u_int16_t proto)
 336 {
 337         struct nf_icmp_net *in = icmp_pernet(net);
 338         struct nf_proto_net *pn = &in->pn;
 339
 340         in->timeout = nf_ct_icmp_timeout;
 341
 342         return icmp_kmemdup_sysctl_table(pn, in);
 343 }
 344
 345 static struct nf_proto_net *icmp_get_net_proto(struct net *net)
 346 {
 347         return &net->ct.nf_ct_proto.icmp.pn;
 348 }
 349
 350 struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp __read_mostly =
 351 {
 352         .l3proto                = PF_INET,
 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,
 358         .new                    = icmp_new,
 359         .error                  = icmp_error,
 360         .destroy                = NULL,
 361         .me                     = NULL,
 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,
 367 #endif
 368 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 369         .ctnl_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,
 375         },
 376 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
 377         .init_net               = icmp_init_net,
 378         .get_net_proto          = icmp_get_net_proto,
 379 };