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                  enum ip_conntrack_info *ctinfo,
 132                  unsigned int hooknum)
 133 {
 134         struct nf_conntrack_tuple innertuple, origtuple;
 135         const struct nf_conntrack_l4proto *innerproto;
 136         const struct nf_conntrack_tuple_hash *h;
 137         const struct nf_conntrack_zone *zone;
 138         struct nf_conntrack_zone tmp;
 139
 140         NF_CT_ASSERT(skb->nfct == NULL);
 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_slow */
 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         skb->nfct = &nf_ct_tuplehash_to_ctrack(h)->ct_general;
 176         skb->nfctinfo = *ctinfo;
 177         return NF_ACCEPT;
 178 }
 179
 180 /* Small and modified version of icmp_rcv */
 181 static int
 182 icmp_error(struct net *net, struct nf_conn *tmpl,
 183            struct sk_buff *skb, unsigned int dataoff,
 184            enum ip_conntrack_info *ctinfo, u_int8_t pf, unsigned int hooknum)
 185 {
 186         const struct icmphdr *icmph;
 187         struct icmphdr _ih;
 188
 189         /* Not enough header? */
 190         icmph = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_ih), &_ih);
 191         if (icmph == NULL) {
 192                 if (LOG_INVALID(net, IPPROTO_ICMP))
 193                         nf_log_packet(net, PF_INET, 0, skb, NULL, NULL,
 194                                       NULL, "nf_ct_icmp: short packet ");
 195                 return -NF_ACCEPT;
 196         }
 197
 198         /* See ip_conntrack_proto_tcp.c */
 199         if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
 200             nf_ip_checksum(skb, hooknum, dataoff, 0)) {
 201                 if (LOG_INVALID(net, IPPROTO_ICMP))
 202                         nf_log_packet(net, PF_INET, 0, skb, NULL, NULL, NULL,
 203                                       "nf_ct_icmp: bad HW ICMP checksum ");
 204                 return -NF_ACCEPT;
 205         }
 206
 207         /*
 208          *      18 is the highest 'known' ICMP type. Anything else is a mystery
 209          *
 210          *      RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently
 211          *                discarded.
 212          */
 213         if (icmph->type > NR_ICMP_TYPES) {
 214                 if (LOG_INVALID(net, IPPROTO_ICMP))
 215                         nf_log_packet(net, PF_INET, 0, skb, NULL, NULL, NULL,
 216                                       "nf_ct_icmp: invalid ICMP type ");
 217                 return -NF_ACCEPT;
 218         }
 219
 220         /* Need to track icmp error message? */
 221         if (icmph->type != ICMP_DEST_UNREACH &&
 222             icmph->type != ICMP_SOURCE_QUENCH &&
 223             icmph->type != ICMP_TIME_EXCEEDED &&
 224             icmph->type != ICMP_PARAMETERPROB &&
 225             icmph->type != ICMP_REDIRECT)
 226                 return NF_ACCEPT;
 227
 228         return icmp_error_message(net, tmpl, skb, ctinfo, hooknum);
 229 }
 230
 231 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 232
 233 #include <linux/netfilter/nfnetlink.h>
 234 #include <linux/netfilter/nfnetlink_conntrack.h>
 235
 236 static int icmp_tuple_to_nlattr(struct sk_buff *skb,
 237                                 const struct nf_conntrack_tuple *t)
 238 {
 239         if (nla_put_be16(skb, CTA_PROTO_ICMP_ID, t->src.u.icmp.id) ||
 240             nla_put_u8(skb, CTA_PROTO_ICMP_TYPE, t->dst.u.icmp.type) ||
 241             nla_put_u8(skb, CTA_PROTO_ICMP_CODE, t->dst.u.icmp.code))
 242                 goto nla_put_failure;
 243         return 0;
 244
 245 nla_put_failure:
 246         return -1;
 247 }
 248
 249 static const struct nla_policy icmp_nla_policy[CTA_PROTO_MAX+1] = {
 250         [CTA_PROTO_ICMP_TYPE]   = { .type = NLA_U8 },
 251         [CTA_PROTO_ICMP_CODE]   = { .type = NLA_U8 },
 252         [CTA_PROTO_ICMP_ID]     = { .type = NLA_U16 },
 253 };
 254
 255 static int icmp_nlattr_to_tuple(struct nlattr *tb[],
 256                                 struct nf_conntrack_tuple *tuple)
 257 {
 258         if (!tb[CTA_PROTO_ICMP_TYPE] ||
 259             !tb[CTA_PROTO_ICMP_CODE] ||
 260             !tb[CTA_PROTO_ICMP_ID])
 261                 return -EINVAL;
 262
 263         tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMP_TYPE]);
 264         tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMP_CODE]);
 265         tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMP_ID]);
 266
 267         if (tuple->dst.u.icmp.type >= sizeof(invmap) ||
 268             !invmap[tuple->dst.u.icmp.type])
 269                 return -EINVAL;
 270
 271         return 0;
 272 }
 273
 274 static int icmp_nlattr_tuple_size(void)
 275 {
 276         return nla_policy_len(icmp_nla_policy, CTA_PROTO_MAX + 1);
 277 }
 278 #endif
 279
 280 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 281
 282 #include <linux/netfilter/nfnetlink.h>
 283 #include <linux/netfilter/nfnetlink_cttimeout.h>
 284
 285 static int icmp_timeout_nlattr_to_obj(struct nlattr *tb[],
 286                                       struct net *net, void *data)
 287 {
 288         unsigned int *timeout = data;
 289         struct nf_icmp_net *in = icmp_pernet(net);
 290
 291         if (tb[CTA_TIMEOUT_ICMP_TIMEOUT]) {
 292                 *timeout =
 293                         ntohl(nla_get_be32(tb[CTA_TIMEOUT_ICMP_TIMEOUT])) * HZ;
 294         } else {
 295                 /* Set default ICMP timeout. */
 296                 *timeout = in->timeout;
 297         }
 298         return 0;
 299 }
 300
 301 static int
 302 icmp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
 303 {
 304         const unsigned int *timeout = data;
 305
 306         if (nla_put_be32(skb, CTA_TIMEOUT_ICMP_TIMEOUT, htonl(*timeout / HZ)))
 307                 goto nla_put_failure;
 308         return 0;
 309
 310 nla_put_failure:
 311         return -ENOSPC;
 312 }
 313
 314 static const struct nla_policy
 315 icmp_timeout_nla_policy[CTA_TIMEOUT_ICMP_MAX+1] = {
 316         [CTA_TIMEOUT_ICMP_TIMEOUT]      = { .type = NLA_U32 },
 317 };
 318 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
 319
 320 #ifdef CONFIG_SYSCTL
 321 static struct ctl_table icmp_sysctl_table[] = {
 322         {
 323                 .procname       = "nf_conntrack_icmp_timeout",
 324                 .maxlen         = sizeof(unsigned int),
 325                 .mode           = 0644,
 326                 .proc_handler   = proc_dointvec_jiffies,
 327         },
 328         { }
 329 };
 330 #endif /* CONFIG_SYSCTL */
 331
 332 static int icmp_kmemdup_sysctl_table(struct nf_proto_net *pn,
 333                                      struct nf_icmp_net *in)
 334 {
 335 #ifdef CONFIG_SYSCTL
 336         pn->ctl_table = kmemdup(icmp_sysctl_table,
 337                                 sizeof(icmp_sysctl_table),
 338                                 GFP_KERNEL);
 339         if (!pn->ctl_table)
 340                 return -ENOMEM;
 341
 342         pn->ctl_table[0].data = &in->timeout;
 343 #endif
 344         return 0;
 345 }
 346
 347 static int icmp_init_net(struct net *net, u_int16_t proto)
 348 {
 349         struct nf_icmp_net *in = icmp_pernet(net);
 350         struct nf_proto_net *pn = &in->pn;
 351
 352         in->timeout = nf_ct_icmp_timeout;
 353
 354         return icmp_kmemdup_sysctl_table(pn, in);
 355 }
 356
 357 static struct nf_proto_net *icmp_get_net_proto(struct net *net)
 358 {
 359         return &net->ct.nf_ct_proto.icmp.pn;
 360 }
 361
 362 struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp __read_mostly =
 363 {
 364         .l3proto                = PF_INET,
 365         .l4proto                = IPPROTO_ICMP,
 366         .name                   = "icmp",
 367         .pkt_to_tuple           = icmp_pkt_to_tuple,
 368         .invert_tuple           = icmp_invert_tuple,
 369         .print_tuple            = icmp_print_tuple,
 370         .packet                 = icmp_packet,
 371         .get_timeouts           = icmp_get_timeouts,
 372         .new                    = icmp_new,
 373         .error                  = icmp_error,
 374         .destroy                = NULL,
 375         .me                     = NULL,
 376 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 377         .tuple_to_nlattr        = icmp_tuple_to_nlattr,
 378         .nlattr_tuple_size      = icmp_nlattr_tuple_size,
 379         .nlattr_to_tuple        = icmp_nlattr_to_tuple,
 380         .nla_policy             = icmp_nla_policy,
 381 #endif
 382 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
 383         .ctnl_timeout           = {
 384                 .nlattr_to_obj  = icmp_timeout_nlattr_to_obj,
 385                 .obj_to_nlattr  = icmp_timeout_obj_to_nlattr,
 386                 .nlattr_max     = CTA_TIMEOUT_ICMP_MAX,
 387                 .obj_size       = sizeof(unsigned int),
 388                 .nla_policy     = icmp_timeout_nla_policy,
 389         },
 390 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
 391         .init_net               = icmp_init_net,
 392         .get_net_proto          = icmp_get_net_proto,
 393 };