net/ipv4/netfilter/ip_conntrack_standalone.c

   1 /* This file contains all the functions required for the standalone
   2    ip_conntrack module.
   3
   4    These are not required by the compatibility layer.
   5 */
   6
   7 /* (C) 1999-2001 Paul `Rusty' Russell
   8  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
   9  *
  10  * This program is free software; you can redistribute it and/or modify
  11  * it under the terms of the GNU General Public License version 2 as
  12  * published by the Free Software Foundation.
  13  */
  14
  15 #include <linux/config.h>
  16 #include <linux/types.h>
  17 #include <linux/ip.h>
  18 #include <linux/netfilter.h>
  19 #include <linux/netfilter_ipv4.h>
  20 #include <linux/module.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/proc_fs.h>
  23 #include <linux/seq_file.h>
  24 #include <linux/percpu.h>
  25 #ifdef CONFIG_SYSCTL
  26 #include <linux/sysctl.h>
  27 #endif
  28 #include <net/checksum.h>
  29 #include <net/ip.h>
  30
  31 #define ASSERT_READ_LOCK(x) MUST_BE_READ_LOCKED(&ip_conntrack_lock)
  32 #define ASSERT_WRITE_LOCK(x) MUST_BE_WRITE_LOCKED(&ip_conntrack_lock)
  33
  34 #include <linux/netfilter_ipv4/ip_conntrack.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  37 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  38 #include <linux/netfilter_ipv4/listhelp.h>
  39
  40 #if 0
  41 #define DEBUGP printk
  42 #else
  43 #define DEBUGP(format, args...)
  44 #endif
  45
  46 MODULE_LICENSE("GPL");
  47
  48 extern atomic_t ip_conntrack_count;
  49 DECLARE_PER_CPU(struct ip_conntrack_stat, ip_conntrack_stat);
  50
  51 static int kill_proto(const struct ip_conntrack *i, void *data)
  52 {
  53         return (i->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum ==
  54                         *((u_int8_t *) data));
  55 }
  56
  57 #ifdef CONFIG_PROC_FS
  58 static int
  59 print_tuple(struct seq_file *s, const struct ip_conntrack_tuple *tuple,
  60             struct ip_conntrack_protocol *proto)
  61 {
  62         seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  63                    NIPQUAD(tuple->src.ip), NIPQUAD(tuple->dst.ip));
  64         return proto->print_tuple(s, tuple);
  65 }
  66
  67 #ifdef CONFIG_IP_NF_CT_ACCT
  68 static unsigned int
  69 seq_print_counters(struct seq_file *s, struct ip_conntrack_counter *counter)
  70 {
  71         return seq_printf(s, "packets=%llu bytes=%llu ",
  72                           (unsigned long long)counter->packets,
  73                           (unsigned long long)counter->bytes);
  74 }
  75 #else
  76 #define seq_print_counters(x, y)        0
  77 #endif
  78
  79 static void *ct_seq_start(struct seq_file *s, loff_t *pos)
  80 {
  81         if (*pos >= ip_conntrack_htable_size)
  82                 return NULL;
  83         return &ip_conntrack_hash[*pos];
  84 }
  85
  86 static void ct_seq_stop(struct seq_file *s, void *v)
  87 {
  88 }
  89
  90 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
  91 {
  92         (*pos)++;
  93         if (*pos >= ip_conntrack_htable_size)
  94                 return NULL;
  95         return &ip_conntrack_hash[*pos];
  96 }
  97
  98 /* return 0 on success, 1 in case of error */
  99 static int ct_seq_real_show(const struct ip_conntrack_tuple_hash *hash,
 100                             struct seq_file *s)
 101 {
 102         struct ip_conntrack *conntrack = hash->ctrack;
 103         struct ip_conntrack_protocol *proto;
 104
 105         MUST_BE_READ_LOCKED(&ip_conntrack_lock);
 106
 107         IP_NF_ASSERT(conntrack);
 108
 109         /* we only want to print DIR_ORIGINAL */
 110         if (DIRECTION(hash))
 111                 return 0;
 112
 113         proto = ip_ct_find_proto(conntrack->tuplehash[IP_CT_DIR_ORIGINAL]
 114                                .tuple.dst.protonum);
 115         IP_NF_ASSERT(proto);
 116
 117         if (seq_printf(s, "%-8s %u %lu ",
 118                       proto->name,
 119                       conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum,
 120                       timer_pending(&conntrack->timeout)
 121                       ? (conntrack->timeout.expires - jiffies)/HZ : 0) != 0)
 122                 return 1;
 123
 124         if (proto->print_conntrack(s, conntrack))
 125                 return 1;
 126
 127         if (print_tuple(s, &conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 128                         proto))
 129                 return 1;
 130
 131         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_ORIGINAL]))
 132                 return 1;
 133
 134         if (!(test_bit(IPS_SEEN_REPLY_BIT, &conntrack->status)))
 135                 if (seq_printf(s, "[UNREPLIED] "))
 136                         return 1;
 137
 138         if (print_tuple(s, &conntrack->tuplehash[IP_CT_DIR_REPLY].tuple,
 139                         proto))
 140                 return 1;
 141
 142         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_REPLY]))
 143                 return 1;
 144
 145         if (test_bit(IPS_ASSURED_BIT, &conntrack->status))
 146                 if (seq_printf(s, "[ASSURED] "))
 147                         return 1;
 148
 149         if (seq_printf(s, "use=%u\n", atomic_read(&conntrack->ct_general.use)))
 150                 return 1;
 151
 152         return 0;
 153 }
 154
 155 static int ct_seq_show(struct seq_file *s, void *v)
 156 {
 157         struct list_head *list = v;
 158         int ret = 0;
 159
 160         /* FIXME: Simply truncates if hash chain too long. */
 161         READ_LOCK(&ip_conntrack_lock);
 162         if (LIST_FIND(list, ct_seq_real_show,
 163                       struct ip_conntrack_tuple_hash *, s))
 164                 ret = -ENOSPC;
 165         READ_UNLOCK(&ip_conntrack_lock);
 166         return ret;
 167 }
 168
 169 static struct seq_operations ct_seq_ops = {
 170         .start = ct_seq_start,
 171         .next  = ct_seq_next,
 172         .stop  = ct_seq_stop,
 173         .show  = ct_seq_show
 174 };
 175
 176 static int ct_open(struct inode *inode, struct file *file)
 177 {
 178         return seq_open(file, &ct_seq_ops);
 179 }
 180
 181 static struct file_operations ct_file_ops = {
 182         .owner   = THIS_MODULE,
 183         .open    = ct_open,
 184         .read    = seq_read,
 185         .llseek  = seq_lseek,
 186         .release = seq_release
 187 };
 188
 189 /* expects */
 190 static void *exp_seq_start(struct seq_file *s, loff_t *pos)
 191 {
 192         struct list_head *e = &ip_conntrack_expect_list;
 193         loff_t i;
 194
 195         /* strange seq_file api calls stop even if we fail,
 196          * thus we need to grab lock since stop unlocks */
 197         READ_LOCK(&ip_conntrack_lock);
 198         READ_LOCK(&ip_conntrack_expect_tuple_lock);
 199
 200         if (list_empty(e))
 201                 return NULL;
 202
 203         for (i = 0; i <= *pos; i++) {
 204                 e = e->next;
 205                 if (e == &ip_conntrack_expect_list)
 206                         return NULL;
 207         }
 208         return e;
 209 }
 210
 211 static void *exp_seq_next(struct seq_file *s, void *v, loff_t *pos)
 212 {
 213         struct list_head *e = v;
 214
 215         e = e->next;
 216
 217         if (e == &ip_conntrack_expect_list)
 218                 return NULL;
 219
 220         return e;
 221 }
 222
 223 static void exp_seq_stop(struct seq_file *s, void *v)
 224 {
 225         READ_UNLOCK(&ip_conntrack_expect_tuple_lock);
 226         READ_UNLOCK(&ip_conntrack_lock);
 227 }
 228
 229 static int exp_seq_show(struct seq_file *s, void *v)
 230 {
 231         struct ip_conntrack_expect *expect = v;
 232
 233         if (expect->expectant->helper->timeout)
 234                 seq_printf(s, "%lu ", timer_pending(&expect->timeout)
 235                            ? (expect->timeout.expires - jiffies)/HZ : 0);
 236         else
 237                 seq_printf(s, "- ");
 238
 239         seq_printf(s, "use=%u proto=%u ", atomic_read(&expect->use),
 240                    expect->tuple.dst.protonum);
 241
 242         print_tuple(s, &expect->tuple,
 243                     ip_ct_find_proto(expect->tuple.dst.protonum));
 244         return seq_putc(s, '\n');
 245 }
 246
 247 static struct seq_operations exp_seq_ops = {
 248         .start = exp_seq_start,
 249         .next = exp_seq_next,
 250         .stop = exp_seq_stop,
 251         .show = exp_seq_show
 252 };
 253
 254 static int exp_open(struct inode *inode, struct file *file)
 255 {
 256         return seq_open(file, &exp_seq_ops);
 257 }
 258
 259 static struct file_operations exp_file_ops = {
 260         .owner   = THIS_MODULE,
 261         .open    = exp_open,
 262         .read    = seq_read,
 263         .llseek  = seq_lseek,
 264         .release = seq_release
 265 };
 266
 267 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 268 {
 269         int cpu;
 270
 271         if (*pos == 0)
 272                 return SEQ_START_TOKEN;
 273
 274         for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
 275                 if (!cpu_possible(cpu))
 276                         continue;
 277                 *pos = cpu+1;
 278                 return &per_cpu(ip_conntrack_stat, cpu);
 279         }
 280
 281         return NULL;
 282 }
 283
 284 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 285 {
 286         int cpu;
 287
 288         for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
 289                 if (!cpu_possible(cpu))
 290                         continue;
 291                 *pos = cpu+1;
 292                 return &per_cpu(ip_conntrack_stat, cpu);
 293         }
 294
 295         return NULL;
 296 }
 297
 298 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 299 {
 300 }
 301
 302 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 303 {
 304         unsigned int nr_conntracks = atomic_read(&ip_conntrack_count);
 305         struct ip_conntrack_stat *st = v;
 306
 307         if (v == SEQ_START_TOKEN) {
 308                 seq_printf(seq, "entries  searched found new invalid ignore delete delete_list insert insert_failed drop early_drop icmp_error  expect_new expect_create expect_delete\n");
 309                 return 0;
 310         }
 311
 312         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 313                         "%08x %08x %08x %08x %08x  %08x %08x %08x \n",
 314                    nr_conntracks,
 315                    st->searched,
 316                    st->found,
 317                    st->new,
 318                    st->invalid,
 319                    st->ignore,
 320                    st->delete,
 321                    st->delete_list,
 322                    st->insert,
 323                    st->insert_failed,
 324                    st->drop,
 325                    st->early_drop,
 326                    st->error,
 327
 328                    st->expect_new,
 329                    st->expect_create,
 330                    st->expect_delete
 331                 );
 332         return 0;
 333 }
 334
 335 static struct seq_operations ct_cpu_seq_ops = {
 336         .start  = ct_cpu_seq_start,
 337         .next   = ct_cpu_seq_next,
 338         .stop   = ct_cpu_seq_stop,
 339         .show   = ct_cpu_seq_show,
 340 };
 341
 342 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 343 {
 344         return seq_open(file, &ct_cpu_seq_ops);
 345 }
 346
 347 static struct file_operations ct_cpu_seq_fops = {
 348         .owner   = THIS_MODULE,
 349         .open    = ct_cpu_seq_open,
 350         .read    = seq_read,
 351         .llseek  = seq_lseek,
 352         .release = seq_release_private,
 353 };
 354 #endif
 355
 356 static unsigned int ip_confirm(unsigned int hooknum,
 357                                struct sk_buff **pskb,
 358                                const struct net_device *in,
 359                                const struct net_device *out,
 360                                int (*okfn)(struct sk_buff *))
 361 {
 362         /* We've seen it coming out the other side: confirm it */
 363         return ip_conntrack_confirm(*pskb);
 364 }
 365
 366 static unsigned int ip_conntrack_defrag(unsigned int hooknum,
 367                                         struct sk_buff **pskb,
 368                                         const struct net_device *in,
 369                                         const struct net_device *out,
 370                                         int (*okfn)(struct sk_buff *))
 371 {
 372         /* Previously seen (loopback)?  Ignore.  Do this before
 373            fragment check. */
 374         if ((*pskb)->nfct)
 375                 return NF_ACCEPT;
 376
 377         /* Gather fragments. */
 378         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 379                 *pskb = ip_ct_gather_frags(*pskb);
 380                 if (!*pskb)
 381                         return NF_STOLEN;
 382         }
 383         return NF_ACCEPT;
 384 }
 385
 386 static unsigned int ip_refrag(unsigned int hooknum,
 387                               struct sk_buff **pskb,
 388                               const struct net_device *in,
 389                               const struct net_device *out,
 390                               int (*okfn)(struct sk_buff *))
 391 {
 392         struct rtable *rt = (struct rtable *)(*pskb)->dst;
 393
 394         /* We've seen it coming out the other side: confirm */
 395         if (ip_confirm(hooknum, pskb, in, out, okfn) != NF_ACCEPT)
 396                 return NF_DROP;
 397
 398         /* Local packets are never produced too large for their
 399            interface.  We degfragment them at LOCAL_OUT, however,
 400            so we have to refragment them here. */
 401         if ((*pskb)->len > dst_pmtu(&rt->u.dst) &&
 402             !skb_shinfo(*pskb)->tso_size) {
 403                 /* No hook can be after us, so this should be OK. */
 404                 ip_fragment(*pskb, okfn);
 405                 return NF_STOLEN;
 406         }
 407         return NF_ACCEPT;
 408 }
 409
 410 static unsigned int ip_conntrack_local(unsigned int hooknum,
 411                                        struct sk_buff **pskb,
 412                                        const struct net_device *in,
 413                                        const struct net_device *out,
 414                                        int (*okfn)(struct sk_buff *))
 415 {
 416         /* root is playing with raw sockets. */
 417         if ((*pskb)->len < sizeof(struct iphdr)
 418             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 419                 if (net_ratelimit())
 420                         printk("ipt_hook: happy cracking.\n");
 421                 return NF_ACCEPT;
 422         }
 423         return ip_conntrack_in(hooknum, pskb, in, out, okfn);
 424 }
 425
 426 /* Connection tracking may drop packets, but never alters them, so
 427    make it the first hook. */
 428 static struct nf_hook_ops ip_conntrack_defrag_ops = {
 429         .hook           = ip_conntrack_defrag,
 430         .owner          = THIS_MODULE,
 431         .pf             = PF_INET,
 432         .hooknum        = NF_IP_PRE_ROUTING,
 433         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 434 };
 435
 436 static struct nf_hook_ops ip_conntrack_in_ops = {
 437         .hook           = ip_conntrack_in,
 438         .owner          = THIS_MODULE,
 439         .pf             = PF_INET,
 440         .hooknum        = NF_IP_PRE_ROUTING,
 441         .priority       = NF_IP_PRI_CONNTRACK,
 442 };
 443
 444 static struct nf_hook_ops ip_conntrack_defrag_local_out_ops = {
 445         .hook           = ip_conntrack_defrag,
 446         .owner          = THIS_MODULE,
 447         .pf             = PF_INET,
 448         .hooknum        = NF_IP_LOCAL_OUT,
 449         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 450 };
 451
 452 static struct nf_hook_ops ip_conntrack_local_out_ops = {
 453         .hook           = ip_conntrack_local,
 454         .owner          = THIS_MODULE,
 455         .pf             = PF_INET,
 456         .hooknum        = NF_IP_LOCAL_OUT,
 457         .priority       = NF_IP_PRI_CONNTRACK,
 458 };
 459
 460 /* Refragmenter; last chance. */
 461 static struct nf_hook_ops ip_conntrack_out_ops = {
 462         .hook           = ip_refrag,
 463         .owner          = THIS_MODULE,
 464         .pf             = PF_INET,
 465         .hooknum        = NF_IP_POST_ROUTING,
 466         .priority       = NF_IP_PRI_LAST,
 467 };
 468
 469 static struct nf_hook_ops ip_conntrack_local_in_ops = {
 470         .hook           = ip_confirm,
 471         .owner          = THIS_MODULE,
 472         .pf             = PF_INET,
 473         .hooknum        = NF_IP_LOCAL_IN,
 474         .priority       = NF_IP_PRI_LAST-1,
 475 };
 476
 477 /* Sysctl support */
 478
 479 #ifdef CONFIG_SYSCTL
 480
 481 /* From ip_conntrack_core.c */
 482 extern int ip_conntrack_max;
 483 extern unsigned int ip_conntrack_htable_size;
 484
 485 /* From ip_conntrack_proto_tcp.c */
 486 extern unsigned long ip_ct_tcp_timeout_syn_sent;
 487 extern unsigned long ip_ct_tcp_timeout_syn_recv;
 488 extern unsigned long ip_ct_tcp_timeout_established;
 489 extern unsigned long ip_ct_tcp_timeout_fin_wait;
 490 extern unsigned long ip_ct_tcp_timeout_close_wait;
 491 extern unsigned long ip_ct_tcp_timeout_last_ack;
 492 extern unsigned long ip_ct_tcp_timeout_time_wait;
 493 extern unsigned long ip_ct_tcp_timeout_close;
 494 extern unsigned long ip_ct_tcp_timeout_max_retrans;
 495 extern int ip_ct_tcp_loose;
 496 extern int ip_ct_tcp_be_liberal;
 497 extern int ip_ct_tcp_max_retrans;
 498
 499 /* From ip_conntrack_proto_udp.c */
 500 extern unsigned long ip_ct_udp_timeout;
 501 extern unsigned long ip_ct_udp_timeout_stream;
 502
 503 /* From ip_conntrack_proto_icmp.c */
 504 extern unsigned long ip_ct_icmp_timeout;
 505
 506 /* From ip_conntrack_proto_icmp.c */
 507 extern unsigned long ip_ct_generic_timeout;
 508
 509 /* Log invalid packets of a given protocol */
 510 static int log_invalid_proto_min = 0;
 511 static int log_invalid_proto_max = 255;
 512
 513 static struct ctl_table_header *ip_ct_sysctl_header;
 514
 515 static ctl_table ip_ct_sysctl_table[] = {
 516         {
 517                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 518                 .procname       = "ip_conntrack_max",
 519                 .data           = &ip_conntrack_max,
 520                 .maxlen         = sizeof(int),
 521                 .mode           = 0644,
 522                 .proc_handler   = &proc_dointvec,
 523         },
 524         {
 525                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 526                 .procname       = "ip_conntrack_count",
 527                 .data           = &ip_conntrack_count,
 528                 .maxlen         = sizeof(int),
 529                 .mode           = 0444,
 530                 .proc_handler   = &proc_dointvec,
 531         },
 532         {
 533                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 534                 .procname       = "ip_conntrack_buckets",
 535                 .data           = &ip_conntrack_htable_size,
 536                 .maxlen         = sizeof(unsigned int),
 537                 .mode           = 0444,
 538                 .proc_handler   = &proc_dointvec,
 539         },
 540         {
 541                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_SENT,
 542                 .procname       = "ip_conntrack_tcp_timeout_syn_sent",
 543                 .data           = &ip_ct_tcp_timeout_syn_sent,
 544                 .maxlen         = sizeof(unsigned int),
 545                 .mode           = 0644,
 546                 .proc_handler   = &proc_dointvec_jiffies,
 547         },
 548         {
 549                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_RECV,
 550                 .procname       = "ip_conntrack_tcp_timeout_syn_recv",
 551                 .data           = &ip_ct_tcp_timeout_syn_recv,
 552                 .maxlen         = sizeof(unsigned int),
 553                 .mode           = 0644,
 554                 .proc_handler   = &proc_dointvec_jiffies,
 555         },
 556         {
 557                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_ESTABLISHED,
 558                 .procname       = "ip_conntrack_tcp_timeout_established",
 559                 .data           = &ip_ct_tcp_timeout_established,
 560                 .maxlen         = sizeof(unsigned int),
 561                 .mode           = 0644,
 562                 .proc_handler   = &proc_dointvec_jiffies,
 563         },
 564         {
 565                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_FIN_WAIT,
 566                 .procname       = "ip_conntrack_tcp_timeout_fin_wait",
 567                 .data           = &ip_ct_tcp_timeout_fin_wait,
 568                 .maxlen         = sizeof(unsigned int),
 569                 .mode           = 0644,
 570                 .proc_handler   = &proc_dointvec_jiffies,
 571         },
 572         {
 573                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE_WAIT,
 574                 .procname       = "ip_conntrack_tcp_timeout_close_wait",
 575                 .data           = &ip_ct_tcp_timeout_close_wait,
 576                 .maxlen         = sizeof(unsigned int),
 577                 .mode           = 0644,
 578                 .proc_handler   = &proc_dointvec_jiffies,
 579         },
 580         {
 581                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_LAST_ACK,
 582                 .procname       = "ip_conntrack_tcp_timeout_last_ack",
 583                 .data           = &ip_ct_tcp_timeout_last_ack,
 584                 .maxlen         = sizeof(unsigned int),
 585                 .mode           = 0644,
 586                 .proc_handler   = &proc_dointvec_jiffies,
 587         },
 588         {
 589                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_TIME_WAIT,
 590                 .procname       = "ip_conntrack_tcp_timeout_time_wait",
 591                 .data           = &ip_ct_tcp_timeout_time_wait,
 592                 .maxlen         = sizeof(unsigned int),
 593                 .mode           = 0644,
 594                 .proc_handler   = &proc_dointvec_jiffies,
 595         },
 596         {
 597                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE,
 598                 .procname       = "ip_conntrack_tcp_timeout_close",
 599                 .data           = &ip_ct_tcp_timeout_close,
 600                 .maxlen         = sizeof(unsigned int),
 601                 .mode           = 0644,
 602                 .proc_handler   = &proc_dointvec_jiffies,
 603         },
 604         {
 605                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT,
 606                 .procname       = "ip_conntrack_udp_timeout",
 607                 .data           = &ip_ct_udp_timeout,
 608                 .maxlen         = sizeof(unsigned int),
 609                 .mode           = 0644,
 610                 .proc_handler   = &proc_dointvec_jiffies,
 611         },
 612         {
 613                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT_STREAM,
 614                 .procname       = "ip_conntrack_udp_timeout_stream",
 615                 .data           = &ip_ct_udp_timeout_stream,
 616                 .maxlen         = sizeof(unsigned int),
 617                 .mode           = 0644,
 618                 .proc_handler   = &proc_dointvec_jiffies,
 619         },
 620         {
 621                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 622                 .procname       = "ip_conntrack_icmp_timeout",
 623                 .data           = &ip_ct_icmp_timeout,
 624                 .maxlen         = sizeof(unsigned int),
 625                 .mode           = 0644,
 626                 .proc_handler   = &proc_dointvec_jiffies,
 627         },
 628         {
 629                 .ctl_name       = NET_IPV4_NF_CONNTRACK_GENERIC_TIMEOUT,
 630                 .procname       = "ip_conntrack_generic_timeout",
 631                 .data           = &ip_ct_generic_timeout,
 632                 .maxlen         = sizeof(unsigned int),
 633                 .mode           = 0644,
 634                 .proc_handler   = &proc_dointvec_jiffies,
 635         },
 636         {
 637                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 638                 .procname       = "ip_conntrack_log_invalid",
 639                 .data           = &ip_ct_log_invalid,
 640                 .maxlen         = sizeof(unsigned int),
 641                 .mode           = 0644,
 642                 .proc_handler   = &proc_dointvec_minmax,
 643                 .strategy       = &sysctl_intvec,
 644                 .extra1         = &log_invalid_proto_min,
 645                 .extra2         = &log_invalid_proto_max,
 646         },
 647         {
 648                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_MAX_RETRANS,
 649                 .procname       = "ip_conntrack_tcp_timeout_max_retrans",
 650                 .data           = &ip_ct_tcp_timeout_max_retrans,
 651                 .maxlen         = sizeof(unsigned int),
 652                 .mode           = 0644,
 653                 .proc_handler   = &proc_dointvec_jiffies,
 654         },
 655         {
 656                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_LOOSE,
 657                 .procname       = "ip_conntrack_tcp_loose",
 658                 .data           = &ip_ct_tcp_loose,
 659                 .maxlen         = sizeof(unsigned int),
 660                 .mode           = 0644,
 661                 .proc_handler   = &proc_dointvec,
 662         },
 663         {
 664                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_BE_LIBERAL,
 665                 .procname       = "ip_conntrack_tcp_be_liberal",
 666                 .data           = &ip_ct_tcp_be_liberal,
 667                 .maxlen         = sizeof(unsigned int),
 668                 .mode           = 0644,
 669                 .proc_handler   = &proc_dointvec,
 670         },
 671         {
 672                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_MAX_RETRANS,
 673                 .procname       = "ip_conntrack_tcp_max_retrans",
 674                 .data           = &ip_ct_tcp_max_retrans,
 675                 .maxlen         = sizeof(unsigned int),
 676                 .mode           = 0644,
 677                 .proc_handler   = &proc_dointvec,
 678         },
 679         { .ctl_name = 0 }
 680 };
 681
 682 #define NET_IP_CONNTRACK_MAX 2089
 683
 684 static ctl_table ip_ct_netfilter_table[] = {
 685         {
 686                 .ctl_name       = NET_IPV4_NETFILTER,
 687                 .procname       = "netfilter",
 688                 .mode           = 0555,
 689                 .child          = ip_ct_sysctl_table,
 690         },
 691         {
 692                 .ctl_name       = NET_IP_CONNTRACK_MAX,
 693                 .procname       = "ip_conntrack_max",
 694                 .data           = &ip_conntrack_max,
 695                 .maxlen         = sizeof(int),
 696                 .mode           = 0644,
 697                 .proc_handler   = &proc_dointvec
 698         },
 699         { .ctl_name = 0 }
 700 };
 701
 702 static ctl_table ip_ct_ipv4_table[] = {
 703         {
 704                 .ctl_name       = NET_IPV4,
 705                 .procname       = "ipv4",
 706                 .mode           = 0555,
 707                 .child          = ip_ct_netfilter_table,
 708         },
 709         { .ctl_name = 0 }
 710 };
 711
 712 static ctl_table ip_ct_net_table[] = {
 713         {
 714                 .ctl_name       = CTL_NET,
 715                 .procname       = "net",
 716                 .mode           = 0555,
 717                 .child          = ip_ct_ipv4_table,
 718         },
 719         { .ctl_name = 0 }
 720 };
 721
 722 EXPORT_SYMBOL(ip_ct_log_invalid);
 723 #endif /* CONFIG_SYSCTL */
 724
 725 static int init_or_cleanup(int init)
 726 {
 727 #ifdef CONFIG_PROC_FS
 728         struct proc_dir_entry *proc, *proc_exp, *proc_stat;
 729 #endif
 730         int ret = 0;
 731
 732         if (!init) goto cleanup;
 733
 734         ret = ip_conntrack_init();
 735         if (ret < 0)
 736                 goto cleanup_nothing;
 737
 738 #ifdef CONFIG_PROC_FS
 739         proc = proc_net_fops_create("ip_conntrack", 0440, &ct_file_ops);
 740         if (!proc) goto cleanup_init;
 741
 742         proc_exp = proc_net_fops_create("ip_conntrack_expect", 0440,
 743                                         &exp_file_ops);
 744         if (!proc_exp) goto cleanup_proc;
 745
 746         proc_stat = create_proc_entry("ip_conntrack", S_IRUGO, proc_net_stat);
 747         if (!proc_stat)
 748                 goto cleanup_proc_exp;
 749
 750         proc_stat->proc_fops = &ct_cpu_seq_fops;
 751         proc_stat->owner = THIS_MODULE;
 752 #endif
 753
 754         ret = nf_register_hook(&ip_conntrack_defrag_ops);
 755         if (ret < 0) {
 756                 printk("ip_conntrack: can't register pre-routing defrag hook.\n");
 757                 goto cleanup_proc_stat;
 758         }
 759         ret = nf_register_hook(&ip_conntrack_defrag_local_out_ops);
 760         if (ret < 0) {
 761                 printk("ip_conntrack: can't register local_out defrag hook.\n");
 762                 goto cleanup_defragops;
 763         }
 764         ret = nf_register_hook(&ip_conntrack_in_ops);
 765         if (ret < 0) {
 766                 printk("ip_conntrack: can't register pre-routing hook.\n");
 767                 goto cleanup_defraglocalops;
 768         }
 769         ret = nf_register_hook(&ip_conntrack_local_out_ops);
 770         if (ret < 0) {
 771                 printk("ip_conntrack: can't register local out hook.\n");
 772                 goto cleanup_inops;
 773         }
 774         ret = nf_register_hook(&ip_conntrack_out_ops);
 775         if (ret < 0) {
 776                 printk("ip_conntrack: can't register post-routing hook.\n");
 777                 goto cleanup_inandlocalops;
 778         }
 779         ret = nf_register_hook(&ip_conntrack_local_in_ops);
 780         if (ret < 0) {
 781                 printk("ip_conntrack: can't register local in hook.\n");
 782                 goto cleanup_inoutandlocalops;
 783         }
 784 #ifdef CONFIG_SYSCTL
 785         ip_ct_sysctl_header = register_sysctl_table(ip_ct_net_table, 0);
 786         if (ip_ct_sysctl_header == NULL) {
 787                 printk("ip_conntrack: can't register to sysctl.\n");
 788                 goto cleanup;
 789         }
 790 #endif
 791
 792         return ret;
 793
 794  cleanup:
 795 #ifdef CONFIG_SYSCTL
 796         unregister_sysctl_table(ip_ct_sysctl_header);
 797 #endif
 798         nf_unregister_hook(&ip_conntrack_local_in_ops);
 799  cleanup_inoutandlocalops:
 800         nf_unregister_hook(&ip_conntrack_out_ops);
 801  cleanup_inandlocalops:
 802         nf_unregister_hook(&ip_conntrack_local_out_ops);
 803  cleanup_inops:
 804         nf_unregister_hook(&ip_conntrack_in_ops);
 805  cleanup_defraglocalops:
 806         nf_unregister_hook(&ip_conntrack_defrag_local_out_ops);
 807  cleanup_defragops:
 808         /* Frag queues may hold fragments with skb->dst == NULL */
 809         ip_ct_no_defrag = 1;
 810         synchronize_net();
 811         local_bh_disable();
 812         ipfrag_flush();
 813         local_bh_enable();
 814         nf_unregister_hook(&ip_conntrack_defrag_ops);
 815  cleanup_proc_stat:
 816 #ifdef CONFIG_PROC_FS
 817         proc_net_remove("ip_conntrack_stat");
 818 cleanup_proc_exp:
 819         proc_net_remove("ip_conntrack_exp");
 820  cleanup_proc:
 821         proc_net_remove("ip_conntrack");
 822  cleanup_init:
 823 #endif /* CONFIG_PROC_FS */
 824         ip_conntrack_cleanup();
 825  cleanup_nothing:
 826         return ret;
 827 }
 828
 829 /* FIXME: Allow NULL functions and sub in pointers to generic for
 830    them. --RR */
 831 int ip_conntrack_protocol_register(struct ip_conntrack_protocol *proto)
 832 {
 833         int ret = 0;
 834
 835         WRITE_LOCK(&ip_conntrack_lock);
 836         if (ip_ct_protos[proto->proto] != &ip_conntrack_generic_protocol) {
 837                 ret = -EBUSY;
 838                 goto out;
 839         }
 840         ip_ct_protos[proto->proto] = proto;
 841  out:
 842         WRITE_UNLOCK(&ip_conntrack_lock);
 843         return ret;
 844 }
 845
 846 void ip_conntrack_protocol_unregister(struct ip_conntrack_protocol *proto)
 847 {
 848         WRITE_LOCK(&ip_conntrack_lock);
 849         ip_ct_protos[proto->proto] = &ip_conntrack_generic_protocol;
 850         WRITE_UNLOCK(&ip_conntrack_lock);
 851
 852         /* Somebody could be still looking at the proto in bh. */
 853         synchronize_net();
 854
 855         /* Remove all contrack entries for this protocol */
 856         ip_ct_selective_cleanup(kill_proto, &proto->proto);
 857 }
 858
 859 static int __init init(void)
 860 {
 861         return init_or_cleanup(1);
 862 }
 863
 864 static void __exit fini(void)
 865 {
 866         init_or_cleanup(0);
 867 }
 868
 869 module_init(init);
 870 module_exit(fini);
 871
 872 /* Some modules need us, but don't depend directly on any symbol.
 873    They should call this. */
 874 void need_ip_conntrack(void)
 875 {
 876 }
 877
 878 EXPORT_SYMBOL(ip_conntrack_protocol_register);
 879 EXPORT_SYMBOL(ip_conntrack_protocol_unregister);
 880 EXPORT_SYMBOL(invert_tuplepr);
 881 EXPORT_SYMBOL(ip_conntrack_alter_reply);
 882 EXPORT_SYMBOL(ip_conntrack_destroyed);
 883 EXPORT_SYMBOL(need_ip_conntrack);
 884 EXPORT_SYMBOL(ip_conntrack_helper_register);
 885 EXPORT_SYMBOL(ip_conntrack_helper_unregister);
 886 EXPORT_SYMBOL(ip_ct_selective_cleanup);
 887 EXPORT_SYMBOL(ip_ct_refresh_acct);
 888 EXPORT_SYMBOL(ip_ct_protos);
 889 EXPORT_SYMBOL(ip_ct_find_proto);
 890 EXPORT_SYMBOL(ip_ct_find_helper);
 891 EXPORT_SYMBOL(ip_conntrack_expect_alloc);
 892 EXPORT_SYMBOL(ip_conntrack_expect_related);
 893 EXPORT_SYMBOL(ip_conntrack_change_expect);
 894 EXPORT_SYMBOL(ip_conntrack_unexpect_related);
 895 EXPORT_SYMBOL_GPL(ip_conntrack_expect_find_get);
 896 EXPORT_SYMBOL_GPL(ip_conntrack_expect_put);
 897 EXPORT_SYMBOL(ip_conntrack_tuple_taken);
 898 EXPORT_SYMBOL(ip_ct_gather_frags);
 899 EXPORT_SYMBOL(ip_conntrack_htable_size);
 900 EXPORT_SYMBOL(ip_conntrack_expect_list);
 901 EXPORT_SYMBOL(ip_conntrack_lock);
 902 EXPORT_SYMBOL(ip_conntrack_hash);
 903 EXPORT_SYMBOL(ip_conntrack_untracked);
 904 EXPORT_SYMBOL_GPL(ip_conntrack_find_get);
 905 EXPORT_SYMBOL_GPL(ip_conntrack_put);