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 #if 0
  25 #define DEBUGP printk
  26 #else
  27 #define DEBUGP(format, args...)
  28 #endif
  29
  30 static int icmp_pkt_to_tuple(const struct sk_buff *skb,
  31                              unsigned int dataoff,
  32                              struct nf_conntrack_tuple *tuple)
  33 {
  34         struct icmphdr _hdr, *hp;
  35
  36         hp = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
  37         if (hp == NULL)
  38                 return 0;
  39
  40         tuple->dst.u.icmp.type = hp->type;
  41         tuple->src.u.icmp.id = hp->un.echo.id;
  42         tuple->dst.u.icmp.code = hp->code;
  43
  44         return 1;
  45 }
  46
  47 /* Add 1; spaces filled with 0. */
  48 static const u_int8_t invmap[] = {
  49         [ICMP_ECHO] = ICMP_ECHOREPLY + 1,
  50         [ICMP_ECHOREPLY] = ICMP_ECHO + 1,
  51         [ICMP_TIMESTAMP] = ICMP_TIMESTAMPREPLY + 1,
  52         [ICMP_TIMESTAMPREPLY] = ICMP_TIMESTAMP + 1,
  53         [ICMP_INFO_REQUEST] = ICMP_INFO_REPLY + 1,
  54         [ICMP_INFO_REPLY] = ICMP_INFO_REQUEST + 1,
  55         [ICMP_ADDRESS] = ICMP_ADDRESSREPLY + 1,
  56         [ICMP_ADDRESSREPLY] = ICMP_ADDRESS + 1
  57 };
  58
  59 static int icmp_invert_tuple(struct nf_conntrack_tuple *tuple,
  60                              const struct nf_conntrack_tuple *orig)
  61 {
  62         if (orig->dst.u.icmp.type >= sizeof(invmap)
  63             || !invmap[orig->dst.u.icmp.type])
  64                 return 0;
  65
  66         tuple->src.u.icmp.id = orig->src.u.icmp.id;
  67         tuple->dst.u.icmp.type = invmap[orig->dst.u.icmp.type] - 1;
  68         tuple->dst.u.icmp.code = orig->dst.u.icmp.code;
  69         return 1;
  70 }
  71
  72 /* Print out the per-protocol part of the tuple. */
  73 static int icmp_print_tuple(struct seq_file *s,
  74                             const struct nf_conntrack_tuple *tuple)
  75 {
  76         return seq_printf(s, "type=%u code=%u id=%u ",
  77                           tuple->dst.u.icmp.type,
  78                           tuple->dst.u.icmp.code,
  79                           ntohs(tuple->src.u.icmp.id));
  80 }
  81
  82 /* Print out the private part of the conntrack. */
  83 static int icmp_print_conntrack(struct seq_file *s,
  84                                 const struct nf_conn *conntrack)
  85 {
  86         return 0;
  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                        int pf,
  95                        unsigned int hooknum)
  96 {
  97         /* Try to delete connection immediately after all replies:
  98            won't actually vanish as we still have skb, and del_timer
  99            means this will only run once even if count hits zero twice
 100            (theoretically possible with SMP) */
 101         if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) {
 102                 if (atomic_dec_and_test(&ct->proto.icmp.count)
 103                     && del_timer(&ct->timeout))
 104                         ct->timeout.function((unsigned long)ct);
 105         } else {
 106                 atomic_inc(&ct->proto.icmp.count);
 107                 nf_conntrack_event_cache(IPCT_PROTOINFO_VOLATILE, skb);
 108                 nf_ct_refresh_acct(ct, ctinfo, skb, nf_ct_icmp_timeout);
 109         }
 110
 111         return NF_ACCEPT;
 112 }
 113
 114 /* Called when a new connection for this protocol found. */
 115 static int icmp_new(struct nf_conn *conntrack,
 116                     const struct sk_buff *skb, unsigned int dataoff)
 117 {
 118         static const u_int8_t valid_new[] = {
 119                 [ICMP_ECHO] = 1,
 120                 [ICMP_TIMESTAMP] = 1,
 121                 [ICMP_INFO_REQUEST] = 1,
 122                 [ICMP_ADDRESS] = 1
 123         };
 124
 125         if (conntrack->tuplehash[0].tuple.dst.u.icmp.type >= sizeof(valid_new)
 126             || !valid_new[conntrack->tuplehash[0].tuple.dst.u.icmp.type]) {
 127                 /* Can't create a new ICMP `conn' with this. */
 128                 DEBUGP("icmp: can't create new conn with type %u\n",
 129                        conntrack->tuplehash[0].tuple.dst.u.icmp.type);
 130                 NF_CT_DUMP_TUPLE(&conntrack->tuplehash[0].tuple);
 131                 return 0;
 132         }
 133         atomic_set(&conntrack->proto.icmp.count, 0);
 134         return 1;
 135 }
 136
 137 extern struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4;
 138 /* Returns conntrack if it dealt with ICMP, and filled in skb fields */
 139 static int
 140 icmp_error_message(struct sk_buff *skb,
 141                  enum ip_conntrack_info *ctinfo,
 142                  unsigned int hooknum)
 143 {
 144         struct nf_conntrack_tuple innertuple, origtuple;
 145         struct {
 146                 struct icmphdr icmp;
 147                 struct iphdr ip;
 148         } _in, *inside;
 149         struct nf_conntrack_l4proto *innerproto;
 150         struct nf_conntrack_tuple_hash *h;
 151         int dataoff;
 152
 153         NF_CT_ASSERT(skb->nfct == NULL);
 154
 155         /* Not enough header? */
 156         inside = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_in), &_in);
 157         if (inside == NULL)
 158                 return -NF_ACCEPT;
 159
 160         /* Ignore ICMP's containing fragments (shouldn't happen) */
 161         if (inside->ip.frag_off & htons(IP_OFFSET)) {
 162                 DEBUGP("icmp_error_message: fragment of proto %u\n",
 163                        inside->ip.protocol);
 164                 return -NF_ACCEPT;
 165         }
 166
 167         /* rcu_read_lock()ed by nf_hook_slow */
 168         innerproto = __nf_ct_l4proto_find(PF_INET, inside->ip.protocol);
 169
 170         dataoff = ip_hdrlen(skb) + sizeof(inside->icmp);
 171         /* Are they talking about one of our connections? */
 172         if (!nf_ct_get_tuple(skb, dataoff, dataoff + inside->ip.ihl*4, PF_INET,
 173                              inside->ip.protocol, &origtuple,
 174                              &nf_conntrack_l3proto_ipv4, innerproto)) {
 175                 DEBUGP("icmp_error_message: ! get_tuple p=%u",
 176                        inside->ip.protocol);
 177                 return -NF_ACCEPT;
 178         }
 179
 180         /* Ordinarily, we'd expect the inverted tupleproto, but it's
 181            been preserved inside the ICMP. */
 182         if (!nf_ct_invert_tuple(&innertuple, &origtuple,
 183                                 &nf_conntrack_l3proto_ipv4, innerproto)) {
 184                 DEBUGP("icmp_error_message: no match\n");
 185                 return -NF_ACCEPT;
 186         }
 187
 188         *ctinfo = IP_CT_RELATED;
 189
 190         h = nf_conntrack_find_get(&innertuple);
 191         if (!h) {
 192                 /* Locally generated ICMPs will match inverted if they
 193                    haven't been SNAT'ed yet */
 194                 /* FIXME: NAT code has to handle half-done double NAT --RR */
 195                 if (hooknum == NF_IP_LOCAL_OUT)
 196                         h = nf_conntrack_find_get(&origtuple);
 197
 198                 if (!h) {
 199                         DEBUGP("icmp_error_message: no match\n");
 200                         return -NF_ACCEPT;
 201                 }
 202
 203                 /* Reverse direction from that found */
 204                 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
 205                         *ctinfo += IP_CT_IS_REPLY;
 206         } else {
 207                 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
 208                         *ctinfo += IP_CT_IS_REPLY;
 209         }
 210
 211         /* Update skb to refer to this connection */
 212         skb->nfct = &nf_ct_tuplehash_to_ctrack(h)->ct_general;
 213         skb->nfctinfo = *ctinfo;
 214         return -NF_ACCEPT;
 215 }
 216
 217 /* Small and modified version of icmp_rcv */
 218 static int
 219 icmp_error(struct sk_buff *skb, unsigned int dataoff,
 220            enum ip_conntrack_info *ctinfo, int pf, unsigned int hooknum)
 221 {
 222         struct icmphdr _ih, *icmph;
 223
 224         /* Not enough header? */
 225         icmph = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_ih), &_ih);
 226         if (icmph == NULL) {
 227                 if (LOG_INVALID(IPPROTO_ICMP))
 228                         nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL,
 229                                       "nf_ct_icmp: short packet ");
 230                 return -NF_ACCEPT;
 231         }
 232
 233         /* See ip_conntrack_proto_tcp.c */
 234         if (nf_conntrack_checksum && hooknum == NF_IP_PRE_ROUTING &&
 235             nf_ip_checksum(skb, hooknum, dataoff, 0)) {
 236                 if (LOG_INVALID(IPPROTO_ICMP))
 237                         nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL,
 238                                       "nf_ct_icmp: bad HW ICMP checksum ");
 239                 return -NF_ACCEPT;
 240         }
 241
 242         /*
 243          *      18 is the highest 'known' ICMP type. Anything else is a mystery
 244          *
 245          *      RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently
 246          *                discarded.
 247          */
 248         if (icmph->type > NR_ICMP_TYPES) {
 249                 if (LOG_INVALID(IPPROTO_ICMP))
 250                         nf_log_packet(PF_INET, 0, skb, NULL, NULL, NULL,
 251                                       "nf_ct_icmp: invalid ICMP type ");
 252                 return -NF_ACCEPT;
 253         }
 254
 255         /* Need to track icmp error message? */
 256         if (icmph->type != ICMP_DEST_UNREACH
 257             && icmph->type != ICMP_SOURCE_QUENCH
 258             && icmph->type != ICMP_TIME_EXCEEDED
 259             && icmph->type != ICMP_PARAMETERPROB
 260             && icmph->type != ICMP_REDIRECT)
 261                 return NF_ACCEPT;
 262
 263         return icmp_error_message(skb, ctinfo, hooknum);
 264 }
 265
 266 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 267
 268 #include <linux/netfilter/nfnetlink.h>
 269 #include <linux/netfilter/nfnetlink_conntrack.h>
 270
 271 static int icmp_tuple_to_nfattr(struct sk_buff *skb,
 272                                 const struct nf_conntrack_tuple *t)
 273 {
 274         NFA_PUT(skb, CTA_PROTO_ICMP_ID, sizeof(u_int16_t),
 275                 &t->src.u.icmp.id);
 276         NFA_PUT(skb, CTA_PROTO_ICMP_TYPE, sizeof(u_int8_t),
 277                 &t->dst.u.icmp.type);
 278         NFA_PUT(skb, CTA_PROTO_ICMP_CODE, sizeof(u_int8_t),
 279                 &t->dst.u.icmp.code);
 280
 281         return 0;
 282
 283 nfattr_failure:
 284         return -1;
 285 }
 286
 287 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 288         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 289         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 290         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t)
 291 };
 292
 293 static int icmp_nfattr_to_tuple(struct nfattr *tb[],
 294                                 struct nf_conntrack_tuple *tuple)
 295 {
 296         if (!tb[CTA_PROTO_ICMP_TYPE-1]
 297             || !tb[CTA_PROTO_ICMP_CODE-1]
 298             || !tb[CTA_PROTO_ICMP_ID-1])
 299                 return -EINVAL;
 300
 301         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 302                 return -EINVAL;
 303
 304         tuple->dst.u.icmp.type =
 305                         *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_ICMP_TYPE-1]);
 306         tuple->dst.u.icmp.code =
 307                         *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_ICMP_CODE-1]);
 308         tuple->src.u.icmp.id =
 309                         *(__be16 *)NFA_DATA(tb[CTA_PROTO_ICMP_ID-1]);
 310
 311         if (tuple->dst.u.icmp.type >= sizeof(invmap)
 312             || !invmap[tuple->dst.u.icmp.type])
 313                 return -EINVAL;
 314
 315         return 0;
 316 }
 317 #endif
 318
 319 #ifdef CONFIG_SYSCTL
 320 static struct ctl_table_header *icmp_sysctl_header;
 321 static struct ctl_table icmp_sysctl_table[] = {
 322         {
 323                 .ctl_name       = NET_NF_CONNTRACK_ICMP_TIMEOUT,
 324                 .procname       = "nf_conntrack_icmp_timeout",
 325                 .data           = &nf_ct_icmp_timeout,
 326                 .maxlen         = sizeof(unsigned int),
 327                 .mode           = 0644,
 328                 .proc_handler   = &proc_dointvec_jiffies,
 329         },
 330         {
 331                 .ctl_name = 0
 332         }
 333 };
 334 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
 335 static struct ctl_table icmp_compat_sysctl_table[] = {
 336         {
 337                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 338                 .procname       = "ip_conntrack_icmp_timeout",
 339                 .data           = &nf_ct_icmp_timeout,
 340                 .maxlen         = sizeof(unsigned int),
 341                 .mode           = 0644,
 342                 .proc_handler   = &proc_dointvec_jiffies,
 343         },
 344         {
 345                 .ctl_name = 0
 346         }
 347 };
 348 #endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
 349 #endif /* CONFIG_SYSCTL */
 350
 351 struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp =
 352 {
 353         .l3proto                = PF_INET,
 354         .l4proto                = IPPROTO_ICMP,
 355         .name                   = "icmp",
 356         .pkt_to_tuple           = icmp_pkt_to_tuple,
 357         .invert_tuple           = icmp_invert_tuple,
 358         .print_tuple            = icmp_print_tuple,
 359         .print_conntrack        = icmp_print_conntrack,
 360         .packet                 = icmp_packet,
 361         .new                    = icmp_new,
 362         .error                  = icmp_error,
 363         .destroy                = NULL,
 364         .me                     = NULL,
 365 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 366         .tuple_to_nfattr        = icmp_tuple_to_nfattr,
 367         .nfattr_to_tuple        = icmp_nfattr_to_tuple,
 368 #endif
 369 #ifdef CONFIG_SYSCTL
 370         .ctl_table_header       = &icmp_sysctl_header,
 371         .ctl_table              = icmp_sysctl_table,
 372 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
 373         .ctl_compat_table       = icmp_compat_sysctl_table,
 374 #endif
 375 #endif
 376 };
 377 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_icmp);