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 /* Print out the per-protocol part of the tuple. */
  75 static void icmp_print_tuple(struct seq_file *s,
  76                             const struct nf_conntrack_tuple *tuple)
  77 {
  78         seq_printf(s, "type=%u code=%u id=%u ",
  79                    tuple->dst.u.icmp.type,
  80                    tuple->dst.u.icmp.code,
  81                    ntohs(tuple->src.u.icmp.id));
  82 }
  83
  84 static unsigned int *icmp_get_timeouts(struct net *net)
  85 {
  86         return &icmp_pernet(net)->timeout;
  87 }
  88
  89 /* Returns verdict for packet, or -1 for invalid. */
  90 static int icmp_packet(struct nf_conn *ct,
  91                        const struct sk_buff *skb,
  92                        unsigned int dataoff,
  93                        enum ip_conntrack_info ctinfo,
  94                        u_int8_t pf,
  95                        unsigned int hooknum,
  96                        unsigned int *timeout)
  97 {
  98         /* Do not immediately delete the connection after the first
  99            successful reply to avoid excessive conntrackd traffic
 100            and also to handle correctly ICMP echo reply duplicates. */
 101         nf_ct_refresh_acct(ct, ctinfo, skb, *timeout);
 102
 103         return NF_ACCEPT;
 104 }
 105
 106 /* Called when a new connection for this protocol found. */
 107 static bool icmp_new(struct nf_conn *ct, const struct sk_buff *skb,
 108                      unsigned int dataoff, unsigned int *timeouts)
 109 {
 110         static const u_int8_t valid_new[] = {
 111                 [ICMP_ECHO] = 1,
 112                 [ICMP_TIMESTAMP] = 1,
 113                 [ICMP_INFO_REQUEST] = 1,
 114                 [ICMP_ADDRESS] = 1
 115         };
 116
 117         if (ct->tuplehash[0].tuple.dst.u.icmp.type >= sizeof(valid_new) ||
 118             !valid_new[ct->tuplehash[0].tuple.dst.u.icmp.type]) {
 119                 /* Can't create a new ICMP `conn' with this. */
 120                 pr_debug("icmp: can't create new conn with type %u\n",
 121                          ct->tuplehash[0].tuple.dst.u.icmp.type);
 122                 nf_ct_dump_tuple_ip(&ct->tuplehash[0].tuple);
 123                 return false;
 124         }
 125         return true;
 126 }
 127
 128 /* Returns conntrack if it dealt with ICMP, and filled in skb fields */
 129 static int
 130 icmp_error_message(struct net *net, struct nf_conn *tmpl, struct sk_buff *skb,
 131                  unsigned int hooknum)
 132 {
 133         struct nf_conntrack_tuple innertuple, origtuple;
 134         const struct nf_conntrack_l4proto *innerproto;
 135         const struct nf_conntrack_tuple_hash *h;
 136         const struct nf_conntrack_zone *zone;
 137         enum ip_conntrack_info ctinfo;
 138         struct nf_conntrack_zone tmp;
 139
 140         NF_CT_ASSERT(!skb_nfct(skb));
 141         zone = nf_ct_zone_tmpl(tmpl, skb, &tmp);
 142
 143         /* Are they talking about one of our connections? */
 144         if (!nf_ct_get_tuplepr(skb,
 145                                skb_network_offset(skb) + ip_hdrlen(skb)
 146                                                        + sizeof(struct icmphdr),
 147                                PF_INET, net, &origtuple)) {
 148                 pr_debug("icmp_error_message: failed to get tuple\n");
 149                 return -NF_ACCEPT;
 150         }
 151
 152         /* rcu_read_lock()ed by nf_hook_thresh */
 153         innerproto = __nf_ct_l4proto_find(PF_INET, origtuple.dst.protonum);
 154
 155         /* Ordinarily, we'd expect the inverted tupleproto, but it's
 156            been preserved inside the ICMP. */
 157         if (!nf_ct_invert_tuple(&innertuple, &origtuple,
 158                                 &nf_conntrack_l3proto_ipv4, innerproto)) {
 159                 pr_debug("icmp_error_message: no match\n");
 160                 return -NF_ACCEPT;
 161         }
 162
 163         ctinfo = IP_CT_RELATED;
 164
 165         h = nf_conntrack_find_get(net, zone, &innertuple);
 166         if (!h) {
 167                 pr_debug("icmp_error_message: no match\n");
 168                 return -NF_ACCEPT;
 169         }
 170
 171         if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
 172                 ctinfo += IP_CT_IS_REPLY;
 173
 174         /* Update skb to refer to this connection */
 175         nf_ct_set(skb, nf_ct_tuplehash_to_ctrack(h), ctinfo);
 176         return NF_ACCEPT;
 177 }
 178
 179 /* Small and modified version of icmp_rcv */
 180 static int
 181 icmp_error(struct net *net, struct nf_conn *tmpl,
 182            struct sk_buff *skb, unsigned int dataoff,
 183            u8 pf, unsigned int hooknum)
 184 {
 185         const struct icmphdr *icmph;
 186         struct icmphdr _ih;
 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(net, IPPROTO_ICMP))
 192                         nf_log_packet(net, PF_INET, 0, skb, NULL, NULL,
 193                                       NULL, "nf_ct_icmp: short packet ");
 194                 return -NF_ACCEPT;
 195         }
 196
 197         /* See ip_conntrack_proto_tcp.c */
 198         if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
 199             nf_ip_checksum(skb, hooknum, dataoff, 0)) {
 200                 if (LOG_INVALID(net, IPPROTO_ICMP))
 201                         nf_log_packet(net, 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(net, IPPROTO_ICMP))
 214                         nf_log_packet(net, 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(net, tmpl, skb, hooknum);
 228 }
 229
 230 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 231
 232 #include <linux/netfilter/nfnetlink.h>
 233 #include <linux/netfilter/nfnetlink_conntrack.h>
 234
 235 static int icmp_tuple_to_nlattr(struct sk_buff *skb,
 236                                 const struct nf_conntrack_tuple *t)
 237 {
 238         if (nla_put_be16(skb, CTA_PROTO_ICMP_ID, t->src.u.icmp.id) ||
 239             nla_put_u8(skb, CTA_PROTO_ICMP_TYPE, t->dst.u.icmp.type) ||
 240             nla_put_u8(skb, CTA_PROTO_ICMP_CODE, t->dst.u.icmp.code))
 241                 goto nla_put_failure;
 242         return 0;
 243
 244 nla_put_failure:
 245         return -1;
 246 }
 247
 248 static const struct nla_policy icmp_nla_policy[CTA_PROTO_MAX+1] = {
 249         [CTA_PROTO_ICMP_TYPE]   = { .type = NLA_U8 },
 250         [CTA_PROTO_ICMP_CODE]   = { .type = NLA_U8 },
 251         [CTA_PROTO_ICMP_ID]     = { .type = NLA_U16 },
 252 };
 253
 254 static int icmp_nlattr_to_tuple(struct nlattr *tb[],
 255                                 struct nf_conntrack_tuple *tuple)
 256 {
 257         if (!tb[CTA_PROTO_ICMP_TYPE] ||
 258             !tb[CTA_PROTO_ICMP_CODE] ||
 259             !tb[CTA_PROTO_ICMP_ID])
 260                 return -EINVAL;
 261
 262         tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMP_TYPE]);
 263         tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMP_CODE]);
 264         tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMP_ID]);
 265
 266         if (tuple->dst.u.icmp.type >= sizeof(invmap) ||
 267             !invmap[tuple->dst.u.icmp.type])
 268                 return -EINVAL;
 269
 270         return 0;
 271 }
 272
 273 static int icmp_nlattr_tuple_size(void)
 274 {
 275         return nla_policy_len(icmp_nla_policy, CTA_PROTO_MAX + 1);
 276 }
 277 #endif
 278
 279 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 280
 281 #include <linux/netfilter/nfnetlink.h>
 282 #include <linux/netfilter/nfnetlink_cttimeout.h>
 283
 284 static int icmp_timeout_nlattr_to_obj(struct nlattr *tb[],
 285                                       struct net *net, void *data)
 286 {
 287         unsigned int *timeout = data;
 288         struct nf_icmp_net *in = icmp_pernet(net);
 289
 290         if (tb[CTA_TIMEOUT_ICMP_TIMEOUT]) {
 291                 *timeout =
 292                         ntohl(nla_get_be32(tb[CTA_TIMEOUT_ICMP_TIMEOUT])) * HZ;
 293         } else {
 294                 /* Set default ICMP timeout. */
 295                 *timeout = in->timeout;
 296         }
 297         return 0;
 298 }
 299
 300 static int
 301 icmp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
 302 {
 303         const unsigned int *timeout = data;
 304
 305         if (nla_put_be32(skb, CTA_TIMEOUT_ICMP_TIMEOUT, htonl(*timeout / HZ)))
 306                 goto nla_put_failure;
 307         return 0;
 308
 309 nla_put_failure:
 310         return -ENOSPC;
 311 }
 312
 313 static const struct nla_policy
 314 icmp_timeout_nla_policy[CTA_TIMEOUT_ICMP_MAX+1] = {
 315         [CTA_TIMEOUT_ICMP_TIMEOUT]      = { .type = NLA_U32 },
 316 };
 317 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
 318
 319 #ifdef CONFIG_SYSCTL
 320 static struct ctl_table icmp_sysctl_table[] = {
 321         {
 322                 .procname       = "nf_conntrack_icmp_timeout",
 323                 .maxlen         = sizeof(unsigned int),
 324                 .mode           = 0644,
 325                 .proc_handler   = proc_dointvec_jiffies,
 326         },
 327         { }
 328 };
 329 #endif /* CONFIG_SYSCTL */
 330
 331 static int icmp_kmemdup_sysctl_table(struct nf_proto_net *pn,
 332                                      struct nf_icmp_net *in)
 333 {
 334 #ifdef CONFIG_SYSCTL
 335         pn->ctl_table = kmemdup(icmp_sysctl_table,
 336                                 sizeof(icmp_sysctl_table),
 337                                 GFP_KERNEL);
 338         if (!pn->ctl_table)
 339                 return -ENOMEM;
 340
 341         pn->ctl_table[0].data = &in->timeout;
 342 #endif
 343         return 0;
 344 }
 345
 346 static int icmp_init_net(struct net *net, u_int16_t proto)
 347 {
 348         struct nf_icmp_net *in = icmp_pernet(net);
 349         struct nf_proto_net *pn = &in->pn;
 350
 351         in->timeout = nf_ct_icmp_timeout;
 352
 353         return icmp_kmemdup_sysctl_table(pn, in);
 354 }
 355
 356 static struct nf_proto_net *icmp_get_net_proto(struct net *net)
 357 {
 358         return &net->ct.nf_ct_proto.icmp.pn;
 359 }
 360
 361 struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp __read_mostly =
 362 {
 363         .l3proto                = PF_INET,
 364         .l4proto                = IPPROTO_ICMP,
 365         .name                   = "icmp",
 366         .pkt_to_tuple           = icmp_pkt_to_tuple,
 367         .invert_tuple           = icmp_invert_tuple,
 368         .print_tuple            = icmp_print_tuple,
 369         .packet                 = icmp_packet,
 370         .get_timeouts           = icmp_get_timeouts,
 371         .new                    = icmp_new,
 372         .error                  = icmp_error,
 373         .destroy                = NULL,
 374         .me                     = NULL,
 375 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 376         .tuple_to_nlattr        = icmp_tuple_to_nlattr,
 377         .nlattr_tuple_size      = icmp_nlattr_tuple_size,
 378         .nlattr_to_tuple        = icmp_nlattr_to_tuple,
 379         .nla_policy             = icmp_nla_policy,
 380 #endif
 381 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 382         .ctnl_timeout           = {
 383                 .nlattr_to_obj  = icmp_timeout_nlattr_to_obj,
 384                 .obj_to_nlattr  = icmp_timeout_obj_to_nlattr,
 385                 .nlattr_max     = CTA_TIMEOUT_ICMP_MAX,
 386                 .obj_size       = sizeof(unsigned int),
 387                 .nla_policy     = icmp_timeout_nla_policy,
 388         },
 389 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
 390         .init_net               = icmp_init_net,
 391         .get_net_proto          = icmp_get_net_proto,
 392 };