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(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,
  70                    const struct ip_conntrack_counter *counter)
  71 {
  72         return seq_printf(s, "packets=%llu bytes=%llu ",
  73                           (unsigned long long)counter->packets,
  74                           (unsigned long long)counter->bytes);
  75 }
  76 #else
  77 #define seq_print_counters(x, y)        0
  78 #endif
  79
  80 struct ct_iter_state {
  81         unsigned int bucket;
  82 };
  83
  84 static struct list_head *ct_get_first(struct seq_file *seq)
  85 {
  86         struct ct_iter_state *st = seq->private;
  87
  88         for (st->bucket = 0;
  89              st->bucket < ip_conntrack_htable_size;
  90              st->bucket++) {
  91                 if (!list_empty(&ip_conntrack_hash[st->bucket]))
  92                         return ip_conntrack_hash[st->bucket].next;
  93         }
  94         return NULL;
  95 }
  96
  97 static struct list_head *ct_get_next(struct seq_file *seq, struct list_head *head)
  98 {
  99         struct ct_iter_state *st = seq->private;
 100
 101         head = head->next;
 102         while (head == &ip_conntrack_hash[st->bucket]) {
 103                 if (++st->bucket >= ip_conntrack_htable_size)
 104                         return NULL;
 105                 head = ip_conntrack_hash[st->bucket].next;
 106         }
 107         return head;
 108 }
 109
 110 static struct list_head *ct_get_idx(struct seq_file *seq, loff_t pos)
 111 {
 112         struct list_head *head = ct_get_first(seq);
 113
 114         if (head)
 115                 while (pos && (head = ct_get_next(seq, head)))
 116                         pos--;
 117         return pos ? NULL : head;
 118 }
 119
 120 static void *ct_seq_start(struct seq_file *seq, loff_t *pos)
 121 {
 122         READ_LOCK(&ip_conntrack_lock);
 123         return ct_get_idx(seq, *pos);
 124 }
 125
 126 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
 127 {
 128         (*pos)++;
 129         return ct_get_next(s, v);
 130 }
 131
 132 static void ct_seq_stop(struct seq_file *s, void *v)
 133 {
 134         READ_UNLOCK(&ip_conntrack_lock);
 135 }
 136
 137 static int ct_seq_show(struct seq_file *s, void *v)
 138 {
 139         const struct ip_conntrack_tuple_hash *hash = v;
 140         const struct ip_conntrack *conntrack = tuplehash_to_ctrack(hash);
 141         struct ip_conntrack_protocol *proto;
 142
 143         MUST_BE_READ_LOCKED(&ip_conntrack_lock);
 144         IP_NF_ASSERT(conntrack);
 145
 146         /* we only want to print DIR_ORIGINAL */
 147         if (DIRECTION(hash))
 148                 return 0;
 149
 150         proto = ip_ct_find_proto(conntrack->tuplehash[IP_CT_DIR_ORIGINAL]
 151                                .tuple.dst.protonum);
 152         IP_NF_ASSERT(proto);
 153
 154         if (seq_printf(s, "%-8s %u %ld ",
 155                       proto->name,
 156                       conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum,
 157                       timer_pending(&conntrack->timeout)
 158                       ? (long)(conntrack->timeout.expires - jiffies)/HZ
 159                       : 0) != 0)
 160                 return -ENOSPC;
 161
 162         if (proto->print_conntrack(s, conntrack))
 163                 return -ENOSPC;
 164
 165         if (print_tuple(s, &conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 166                         proto))
 167                 return -ENOSPC;
 168
 169         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_ORIGINAL]))
 170                 return -ENOSPC;
 171
 172         if (!(test_bit(IPS_SEEN_REPLY_BIT, &conntrack->status)))
 173                 if (seq_printf(s, "[UNREPLIED] "))
 174                         return -ENOSPC;
 175
 176         if (print_tuple(s, &conntrack->tuplehash[IP_CT_DIR_REPLY].tuple,
 177                         proto))
 178                 return -ENOSPC;
 179
 180         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_REPLY]))
 181                 return -ENOSPC;
 182
 183         if (test_bit(IPS_ASSURED_BIT, &conntrack->status))
 184                 if (seq_printf(s, "[ASSURED] "))
 185                         return -ENOSPC;
 186
 187 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 188         if (seq_printf(s, "mark=%lu ", conntrack->mark))
 189                 return -ENOSPC;
 190 #endif
 191
 192         if (seq_printf(s, "use=%u\n", atomic_read(&conntrack->ct_general.use)))
 193                 return -ENOSPC;
 194
 195         return 0;
 196 }
 197
 198 static struct seq_operations ct_seq_ops = {
 199         .start = ct_seq_start,
 200         .next  = ct_seq_next,
 201         .stop  = ct_seq_stop,
 202         .show  = ct_seq_show
 203 };
 204
 205 static int ct_open(struct inode *inode, struct file *file)
 206 {
 207         struct seq_file *seq;
 208         struct ct_iter_state *st;
 209         int ret;
 210
 211         st = kmalloc(sizeof(struct ct_iter_state), GFP_KERNEL);
 212         if (st == NULL)
 213                 return -ENOMEM;
 214         ret = seq_open(file, &ct_seq_ops);
 215         if (ret)
 216                 goto out_free;
 217         seq          = file->private_data;
 218         seq->private = st;
 219         memset(st, 0, sizeof(struct ct_iter_state));
 220         return ret;
 221 out_free:
 222         kfree(st);
 223         return ret;
 224 }
 225
 226 static struct file_operations ct_file_ops = {
 227         .owner   = THIS_MODULE,
 228         .open    = ct_open,
 229         .read    = seq_read,
 230         .llseek  = seq_lseek,
 231         .release = seq_release_private,
 232 };
 233
 234 /* expects */
 235 static void *exp_seq_start(struct seq_file *s, loff_t *pos)
 236 {
 237         struct list_head *e = &ip_conntrack_expect_list;
 238         loff_t i;
 239
 240         /* strange seq_file api calls stop even if we fail,
 241          * thus we need to grab lock since stop unlocks */
 242         READ_LOCK(&ip_conntrack_lock);
 243
 244         if (list_empty(e))
 245                 return NULL;
 246
 247         for (i = 0; i <= *pos; i++) {
 248                 e = e->next;
 249                 if (e == &ip_conntrack_expect_list)
 250                         return NULL;
 251         }
 252         return e;
 253 }
 254
 255 static void *exp_seq_next(struct seq_file *s, void *v, loff_t *pos)
 256 {
 257         struct list_head *e = v;
 258
 259         ++*pos;
 260         e = e->next;
 261
 262         if (e == &ip_conntrack_expect_list)
 263                 return NULL;
 264
 265         return e;
 266 }
 267
 268 static void exp_seq_stop(struct seq_file *s, void *v)
 269 {
 270         READ_UNLOCK(&ip_conntrack_lock);
 271 }
 272
 273 static int exp_seq_show(struct seq_file *s, void *v)
 274 {
 275         struct ip_conntrack_expect *expect = v;
 276
 277         if (expect->timeout.function)
 278                 seq_printf(s, "%ld ", timer_pending(&expect->timeout)
 279                            ? (long)(expect->timeout.expires - jiffies)/HZ : 0);
 280         else
 281                 seq_printf(s, "- ");
 282
 283         seq_printf(s, "proto=%u ", expect->tuple.dst.protonum);
 284
 285         print_tuple(s, &expect->tuple,
 286                     ip_ct_find_proto(expect->tuple.dst.protonum));
 287         return seq_putc(s, '\n');
 288 }
 289
 290 static struct seq_operations exp_seq_ops = {
 291         .start = exp_seq_start,
 292         .next = exp_seq_next,
 293         .stop = exp_seq_stop,
 294         .show = exp_seq_show
 295 };
 296
 297 static int exp_open(struct inode *inode, struct file *file)
 298 {
 299         return seq_open(file, &exp_seq_ops);
 300 }
 301
 302 static struct file_operations exp_file_ops = {
 303         .owner   = THIS_MODULE,
 304         .open    = exp_open,
 305         .read    = seq_read,
 306         .llseek  = seq_lseek,
 307         .release = seq_release
 308 };
 309
 310 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 311 {
 312         int cpu;
 313
 314         if (*pos == 0)
 315                 return SEQ_START_TOKEN;
 316
 317         for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
 318                 if (!cpu_possible(cpu))
 319                         continue;
 320                 *pos = cpu+1;
 321                 return &per_cpu(ip_conntrack_stat, cpu);
 322         }
 323
 324         return NULL;
 325 }
 326
 327 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 328 {
 329         int cpu;
 330
 331         for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
 332                 if (!cpu_possible(cpu))
 333                         continue;
 334                 *pos = cpu+1;
 335                 return &per_cpu(ip_conntrack_stat, cpu);
 336         }
 337
 338         return NULL;
 339 }
 340
 341 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 342 {
 343 }
 344
 345 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 346 {
 347         unsigned int nr_conntracks = atomic_read(&ip_conntrack_count);
 348         struct ip_conntrack_stat *st = v;
 349
 350         if (v == SEQ_START_TOKEN) {
 351                 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");
 352                 return 0;
 353         }
 354
 355         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 356                         "%08x %08x %08x %08x %08x  %08x %08x %08x \n",
 357                    nr_conntracks,
 358                    st->searched,
 359                    st->found,
 360                    st->new,
 361                    st->invalid,
 362                    st->ignore,
 363                    st->delete,
 364                    st->delete_list,
 365                    st->insert,
 366                    st->insert_failed,
 367                    st->drop,
 368                    st->early_drop,
 369                    st->error,
 370
 371                    st->expect_new,
 372                    st->expect_create,
 373                    st->expect_delete
 374                 );
 375         return 0;
 376 }
 377
 378 static struct seq_operations ct_cpu_seq_ops = {
 379         .start  = ct_cpu_seq_start,
 380         .next   = ct_cpu_seq_next,
 381         .stop   = ct_cpu_seq_stop,
 382         .show   = ct_cpu_seq_show,
 383 };
 384
 385 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 386 {
 387         return seq_open(file, &ct_cpu_seq_ops);
 388 }
 389
 390 static struct file_operations ct_cpu_seq_fops = {
 391         .owner   = THIS_MODULE,
 392         .open    = ct_cpu_seq_open,
 393         .read    = seq_read,
 394         .llseek  = seq_lseek,
 395         .release = seq_release_private,
 396 };
 397 #endif
 398
 399 static unsigned int ip_confirm(unsigned int hooknum,
 400                                struct sk_buff **pskb,
 401                                const struct net_device *in,
 402                                const struct net_device *out,
 403                                int (*okfn)(struct sk_buff *))
 404 {
 405         /* We've seen it coming out the other side: confirm it */
 406         return ip_conntrack_confirm(pskb);
 407 }
 408
 409 static unsigned int ip_conntrack_help(unsigned int hooknum,
 410                                       struct sk_buff **pskb,
 411                                       const struct net_device *in,
 412                                       const struct net_device *out,
 413                                       int (*okfn)(struct sk_buff *))
 414 {
 415         struct ip_conntrack *ct;
 416         enum ip_conntrack_info ctinfo;
 417
 418         /* This is where we call the helper: as the packet goes out. */
 419         ct = ip_conntrack_get(*pskb, &ctinfo);
 420         if (ct && ct->helper) {
 421                 unsigned int ret;
 422                 ret = ct->helper->help(pskb, ct, ctinfo);
 423                 if (ret != NF_ACCEPT)
 424                         return ret;
 425         }
 426         return NF_ACCEPT;
 427 }
 428
 429 static unsigned int ip_conntrack_defrag(unsigned int hooknum,
 430                                         struct sk_buff **pskb,
 431                                         const struct net_device *in,
 432                                         const struct net_device *out,
 433                                         int (*okfn)(struct sk_buff *))
 434 {
 435         /* Gather fragments. */
 436         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 437                 *pskb = ip_ct_gather_frags(*pskb,
 438                                            hooknum == NF_IP_PRE_ROUTING ?
 439                                            IP_DEFRAG_CONNTRACK_IN :
 440                                            IP_DEFRAG_CONNTRACK_OUT);
 441                 if (!*pskb)
 442                         return NF_STOLEN;
 443         }
 444         return NF_ACCEPT;
 445 }
 446
 447 static unsigned int ip_refrag(unsigned int hooknum,
 448                               struct sk_buff **pskb,
 449                               const struct net_device *in,
 450                               const struct net_device *out,
 451                               int (*okfn)(struct sk_buff *))
 452 {
 453         struct rtable *rt = (struct rtable *)(*pskb)->dst;
 454
 455         /* We've seen it coming out the other side: confirm */
 456         if (ip_confirm(hooknum, pskb, in, out, okfn) != NF_ACCEPT)
 457                 return NF_DROP;
 458
 459         /* Local packets are never produced too large for their
 460            interface.  We degfragment them at LOCAL_OUT, however,
 461            so we have to refragment them here. */
 462         if ((*pskb)->len > dst_mtu(&rt->u.dst) &&
 463             !skb_shinfo(*pskb)->tso_size) {
 464                 /* No hook can be after us, so this should be OK. */
 465                 ip_fragment(*pskb, okfn);
 466                 return NF_STOLEN;
 467         }
 468         return NF_ACCEPT;
 469 }
 470
 471 static unsigned int ip_conntrack_local(unsigned int hooknum,
 472                                        struct sk_buff **pskb,
 473                                        const struct net_device *in,
 474                                        const struct net_device *out,
 475                                        int (*okfn)(struct sk_buff *))
 476 {
 477         /* root is playing with raw sockets. */
 478         if ((*pskb)->len < sizeof(struct iphdr)
 479             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 480                 if (net_ratelimit())
 481                         printk("ipt_hook: happy cracking.\n");
 482                 return NF_ACCEPT;
 483         }
 484         return ip_conntrack_in(hooknum, pskb, in, out, okfn);
 485 }
 486
 487 /* Connection tracking may drop packets, but never alters them, so
 488    make it the first hook. */
 489 static struct nf_hook_ops ip_conntrack_defrag_ops = {
 490         .hook           = ip_conntrack_defrag,
 491         .owner          = THIS_MODULE,
 492         .pf             = PF_INET,
 493         .hooknum        = NF_IP_PRE_ROUTING,
 494         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 495 };
 496
 497 static struct nf_hook_ops ip_conntrack_in_ops = {
 498         .hook           = ip_conntrack_in,
 499         .owner          = THIS_MODULE,
 500         .pf             = PF_INET,
 501         .hooknum        = NF_IP_PRE_ROUTING,
 502         .priority       = NF_IP_PRI_CONNTRACK,
 503 };
 504
 505 static struct nf_hook_ops ip_conntrack_defrag_local_out_ops = {
 506         .hook           = ip_conntrack_defrag,
 507         .owner          = THIS_MODULE,
 508         .pf             = PF_INET,
 509         .hooknum        = NF_IP_LOCAL_OUT,
 510         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 511 };
 512
 513 static struct nf_hook_ops ip_conntrack_local_out_ops = {
 514         .hook           = ip_conntrack_local,
 515         .owner          = THIS_MODULE,
 516         .pf             = PF_INET,
 517         .hooknum        = NF_IP_LOCAL_OUT,
 518         .priority       = NF_IP_PRI_CONNTRACK,
 519 };
 520
 521 /* helpers */
 522 static struct nf_hook_ops ip_conntrack_helper_out_ops = {
 523         .hook           = ip_conntrack_help,
 524         .owner          = THIS_MODULE,
 525         .pf             = PF_INET,
 526         .hooknum        = NF_IP_POST_ROUTING,
 527         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 528 };
 529
 530 static struct nf_hook_ops ip_conntrack_helper_in_ops = {
 531         .hook           = ip_conntrack_help,
 532         .owner          = THIS_MODULE,
 533         .pf             = PF_INET,
 534         .hooknum        = NF_IP_LOCAL_IN,
 535         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 536 };
 537
 538 /* Refragmenter; last chance. */
 539 static struct nf_hook_ops ip_conntrack_out_ops = {
 540         .hook           = ip_refrag,
 541         .owner          = THIS_MODULE,
 542         .pf             = PF_INET,
 543         .hooknum        = NF_IP_POST_ROUTING,
 544         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 545 };
 546
 547 static struct nf_hook_ops ip_conntrack_local_in_ops = {
 548         .hook           = ip_confirm,
 549         .owner          = THIS_MODULE,
 550         .pf             = PF_INET,
 551         .hooknum        = NF_IP_LOCAL_IN,
 552         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 553 };
 554
 555 /* Sysctl support */
 556
 557 #ifdef CONFIG_SYSCTL
 558
 559 /* From ip_conntrack_core.c */
 560 extern int ip_conntrack_max;
 561 extern unsigned int ip_conntrack_htable_size;
 562
 563 /* From ip_conntrack_proto_tcp.c */
 564 extern unsigned long ip_ct_tcp_timeout_syn_sent;
 565 extern unsigned long ip_ct_tcp_timeout_syn_recv;
 566 extern unsigned long ip_ct_tcp_timeout_established;
 567 extern unsigned long ip_ct_tcp_timeout_fin_wait;
 568 extern unsigned long ip_ct_tcp_timeout_close_wait;
 569 extern unsigned long ip_ct_tcp_timeout_last_ack;
 570 extern unsigned long ip_ct_tcp_timeout_time_wait;
 571 extern unsigned long ip_ct_tcp_timeout_close;
 572 extern unsigned long ip_ct_tcp_timeout_max_retrans;
 573 extern int ip_ct_tcp_loose;
 574 extern int ip_ct_tcp_be_liberal;
 575 extern int ip_ct_tcp_max_retrans;
 576
 577 /* From ip_conntrack_proto_udp.c */
 578 extern unsigned long ip_ct_udp_timeout;
 579 extern unsigned long ip_ct_udp_timeout_stream;
 580
 581 /* From ip_conntrack_proto_icmp.c */
 582 extern unsigned long ip_ct_icmp_timeout;
 583
 584 /* From ip_conntrack_proto_icmp.c */
 585 extern unsigned long ip_ct_generic_timeout;
 586
 587 /* Log invalid packets of a given protocol */
 588 static int log_invalid_proto_min = 0;
 589 static int log_invalid_proto_max = 255;
 590
 591 static struct ctl_table_header *ip_ct_sysctl_header;
 592
 593 static ctl_table ip_ct_sysctl_table[] = {
 594         {
 595                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 596                 .procname       = "ip_conntrack_max",
 597                 .data           = &ip_conntrack_max,
 598                 .maxlen         = sizeof(int),
 599                 .mode           = 0644,
 600                 .proc_handler   = &proc_dointvec,
 601         },
 602         {
 603                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 604                 .procname       = "ip_conntrack_count",
 605                 .data           = &ip_conntrack_count,
 606                 .maxlen         = sizeof(int),
 607                 .mode           = 0444,
 608                 .proc_handler   = &proc_dointvec,
 609         },
 610         {
 611                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 612                 .procname       = "ip_conntrack_buckets",
 613                 .data           = &ip_conntrack_htable_size,
 614                 .maxlen         = sizeof(unsigned int),
 615                 .mode           = 0444,
 616                 .proc_handler   = &proc_dointvec,
 617         },
 618         {
 619                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_SENT,
 620                 .procname       = "ip_conntrack_tcp_timeout_syn_sent",
 621                 .data           = &ip_ct_tcp_timeout_syn_sent,
 622                 .maxlen         = sizeof(unsigned int),
 623                 .mode           = 0644,
 624                 .proc_handler   = &proc_dointvec_jiffies,
 625         },
 626         {
 627                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_RECV,
 628                 .procname       = "ip_conntrack_tcp_timeout_syn_recv",
 629                 .data           = &ip_ct_tcp_timeout_syn_recv,
 630                 .maxlen         = sizeof(unsigned int),
 631                 .mode           = 0644,
 632                 .proc_handler   = &proc_dointvec_jiffies,
 633         },
 634         {
 635                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_ESTABLISHED,
 636                 .procname       = "ip_conntrack_tcp_timeout_established",
 637                 .data           = &ip_ct_tcp_timeout_established,
 638                 .maxlen         = sizeof(unsigned int),
 639                 .mode           = 0644,
 640                 .proc_handler   = &proc_dointvec_jiffies,
 641         },
 642         {
 643                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_FIN_WAIT,
 644                 .procname       = "ip_conntrack_tcp_timeout_fin_wait",
 645                 .data           = &ip_ct_tcp_timeout_fin_wait,
 646                 .maxlen         = sizeof(unsigned int),
 647                 .mode           = 0644,
 648                 .proc_handler   = &proc_dointvec_jiffies,
 649         },
 650         {
 651                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE_WAIT,
 652                 .procname       = "ip_conntrack_tcp_timeout_close_wait",
 653                 .data           = &ip_ct_tcp_timeout_close_wait,
 654                 .maxlen         = sizeof(unsigned int),
 655                 .mode           = 0644,
 656                 .proc_handler   = &proc_dointvec_jiffies,
 657         },
 658         {
 659                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_LAST_ACK,
 660                 .procname       = "ip_conntrack_tcp_timeout_last_ack",
 661                 .data           = &ip_ct_tcp_timeout_last_ack,
 662                 .maxlen         = sizeof(unsigned int),
 663                 .mode           = 0644,
 664                 .proc_handler   = &proc_dointvec_jiffies,
 665         },
 666         {
 667                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_TIME_WAIT,
 668                 .procname       = "ip_conntrack_tcp_timeout_time_wait",
 669                 .data           = &ip_ct_tcp_timeout_time_wait,
 670                 .maxlen         = sizeof(unsigned int),
 671                 .mode           = 0644,
 672                 .proc_handler   = &proc_dointvec_jiffies,
 673         },
 674         {
 675                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE,
 676                 .procname       = "ip_conntrack_tcp_timeout_close",
 677                 .data           = &ip_ct_tcp_timeout_close,
 678                 .maxlen         = sizeof(unsigned int),
 679                 .mode           = 0644,
 680                 .proc_handler   = &proc_dointvec_jiffies,
 681         },
 682         {
 683                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT,
 684                 .procname       = "ip_conntrack_udp_timeout",
 685                 .data           = &ip_ct_udp_timeout,
 686                 .maxlen         = sizeof(unsigned int),
 687                 .mode           = 0644,
 688                 .proc_handler   = &proc_dointvec_jiffies,
 689         },
 690         {
 691                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT_STREAM,
 692                 .procname       = "ip_conntrack_udp_timeout_stream",
 693                 .data           = &ip_ct_udp_timeout_stream,
 694                 .maxlen         = sizeof(unsigned int),
 695                 .mode           = 0644,
 696                 .proc_handler   = &proc_dointvec_jiffies,
 697         },
 698         {
 699                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 700                 .procname       = "ip_conntrack_icmp_timeout",
 701                 .data           = &ip_ct_icmp_timeout,
 702                 .maxlen         = sizeof(unsigned int),
 703                 .mode           = 0644,
 704                 .proc_handler   = &proc_dointvec_jiffies,
 705         },
 706         {
 707                 .ctl_name       = NET_IPV4_NF_CONNTRACK_GENERIC_TIMEOUT,
 708                 .procname       = "ip_conntrack_generic_timeout",
 709                 .data           = &ip_ct_generic_timeout,
 710                 .maxlen         = sizeof(unsigned int),
 711                 .mode           = 0644,
 712                 .proc_handler   = &proc_dointvec_jiffies,
 713         },
 714         {
 715                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 716                 .procname       = "ip_conntrack_log_invalid",
 717                 .data           = &ip_ct_log_invalid,
 718                 .maxlen         = sizeof(unsigned int),
 719                 .mode           = 0644,
 720                 .proc_handler   = &proc_dointvec_minmax,
 721                 .strategy       = &sysctl_intvec,
 722                 .extra1         = &log_invalid_proto_min,
 723                 .extra2         = &log_invalid_proto_max,
 724         },
 725         {
 726                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_MAX_RETRANS,
 727                 .procname       = "ip_conntrack_tcp_timeout_max_retrans",
 728                 .data           = &ip_ct_tcp_timeout_max_retrans,
 729                 .maxlen         = sizeof(unsigned int),
 730                 .mode           = 0644,
 731                 .proc_handler   = &proc_dointvec_jiffies,
 732         },
 733         {
 734                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_LOOSE,
 735                 .procname       = "ip_conntrack_tcp_loose",
 736                 .data           = &ip_ct_tcp_loose,
 737                 .maxlen         = sizeof(unsigned int),
 738                 .mode           = 0644,
 739                 .proc_handler   = &proc_dointvec,
 740         },
 741         {
 742                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_BE_LIBERAL,
 743                 .procname       = "ip_conntrack_tcp_be_liberal",
 744                 .data           = &ip_ct_tcp_be_liberal,
 745                 .maxlen         = sizeof(unsigned int),
 746                 .mode           = 0644,
 747                 .proc_handler   = &proc_dointvec,
 748         },
 749         {
 750                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_MAX_RETRANS,
 751                 .procname       = "ip_conntrack_tcp_max_retrans",
 752                 .data           = &ip_ct_tcp_max_retrans,
 753                 .maxlen         = sizeof(unsigned int),
 754                 .mode           = 0644,
 755                 .proc_handler   = &proc_dointvec,
 756         },
 757         { .ctl_name = 0 }
 758 };
 759
 760 #define NET_IP_CONNTRACK_MAX 2089
 761
 762 static ctl_table ip_ct_netfilter_table[] = {
 763         {
 764                 .ctl_name       = NET_IPV4_NETFILTER,
 765                 .procname       = "netfilter",
 766                 .mode           = 0555,
 767                 .child          = ip_ct_sysctl_table,
 768         },
 769         {
 770                 .ctl_name       = NET_IP_CONNTRACK_MAX,
 771                 .procname       = "ip_conntrack_max",
 772                 .data           = &ip_conntrack_max,
 773                 .maxlen         = sizeof(int),
 774                 .mode           = 0644,
 775                 .proc_handler   = &proc_dointvec
 776         },
 777         { .ctl_name = 0 }
 778 };
 779
 780 static ctl_table ip_ct_ipv4_table[] = {
 781         {
 782                 .ctl_name       = NET_IPV4,
 783                 .procname       = "ipv4",
 784                 .mode           = 0555,
 785                 .child          = ip_ct_netfilter_table,
 786         },
 787         { .ctl_name = 0 }
 788 };
 789
 790 static ctl_table ip_ct_net_table[] = {
 791         {
 792                 .ctl_name       = CTL_NET,
 793                 .procname       = "net",
 794                 .mode           = 0555,
 795                 .child          = ip_ct_ipv4_table,
 796         },
 797         { .ctl_name = 0 }
 798 };
 799
 800 EXPORT_SYMBOL(ip_ct_log_invalid);
 801 #endif /* CONFIG_SYSCTL */
 802
 803 static int init_or_cleanup(int init)
 804 {
 805 #ifdef CONFIG_PROC_FS
 806         struct proc_dir_entry *proc, *proc_exp, *proc_stat;
 807 #endif
 808         int ret = 0;
 809
 810         if (!init) goto cleanup;
 811
 812         ret = ip_conntrack_init();
 813         if (ret < 0)
 814                 goto cleanup_nothing;
 815
 816 #ifdef CONFIG_PROC_FS
 817         ret = -ENOMEM;
 818         proc = proc_net_fops_create("ip_conntrack", 0440, &ct_file_ops);
 819         if (!proc) goto cleanup_init;
 820
 821         proc_exp = proc_net_fops_create("ip_conntrack_expect", 0440,
 822                                         &exp_file_ops);
 823         if (!proc_exp) goto cleanup_proc;
 824
 825         proc_stat = create_proc_entry("ip_conntrack", S_IRUGO, proc_net_stat);
 826         if (!proc_stat)
 827                 goto cleanup_proc_exp;
 828
 829         proc_stat->proc_fops = &ct_cpu_seq_fops;
 830         proc_stat->owner = THIS_MODULE;
 831 #endif
 832
 833         ret = nf_register_hook(&ip_conntrack_defrag_ops);
 834         if (ret < 0) {
 835                 printk("ip_conntrack: can't register pre-routing defrag hook.\n");
 836                 goto cleanup_proc_stat;
 837         }
 838         ret = nf_register_hook(&ip_conntrack_defrag_local_out_ops);
 839         if (ret < 0) {
 840                 printk("ip_conntrack: can't register local_out defrag hook.\n");
 841                 goto cleanup_defragops;
 842         }
 843         ret = nf_register_hook(&ip_conntrack_in_ops);
 844         if (ret < 0) {
 845                 printk("ip_conntrack: can't register pre-routing hook.\n");
 846                 goto cleanup_defraglocalops;
 847         }
 848         ret = nf_register_hook(&ip_conntrack_local_out_ops);
 849         if (ret < 0) {
 850                 printk("ip_conntrack: can't register local out hook.\n");
 851                 goto cleanup_inops;
 852         }
 853         ret = nf_register_hook(&ip_conntrack_helper_in_ops);
 854         if (ret < 0) {
 855                 printk("ip_conntrack: can't register local in helper hook.\n");
 856                 goto cleanup_inandlocalops;
 857         }
 858         ret = nf_register_hook(&ip_conntrack_helper_out_ops);
 859         if (ret < 0) {
 860                 printk("ip_conntrack: can't register postrouting helper hook.\n");
 861                 goto cleanup_helperinops;
 862         }
 863         ret = nf_register_hook(&ip_conntrack_out_ops);
 864         if (ret < 0) {
 865                 printk("ip_conntrack: can't register post-routing hook.\n");
 866                 goto cleanup_helperoutops;
 867         }
 868         ret = nf_register_hook(&ip_conntrack_local_in_ops);
 869         if (ret < 0) {
 870                 printk("ip_conntrack: can't register local in hook.\n");
 871                 goto cleanup_inoutandlocalops;
 872         }
 873 #ifdef CONFIG_SYSCTL
 874         ip_ct_sysctl_header = register_sysctl_table(ip_ct_net_table, 0);
 875         if (ip_ct_sysctl_header == NULL) {
 876                 printk("ip_conntrack: can't register to sysctl.\n");
 877                 ret = -ENOMEM;
 878                 goto cleanup_localinops;
 879         }
 880 #endif
 881
 882         return ret;
 883
 884  cleanup:
 885 #ifdef CONFIG_SYSCTL
 886         unregister_sysctl_table(ip_ct_sysctl_header);
 887  cleanup_localinops:
 888 #endif
 889         nf_unregister_hook(&ip_conntrack_local_in_ops);
 890  cleanup_inoutandlocalops:
 891         nf_unregister_hook(&ip_conntrack_out_ops);
 892  cleanup_helperoutops:
 893         nf_unregister_hook(&ip_conntrack_helper_out_ops);
 894  cleanup_helperinops:
 895         nf_unregister_hook(&ip_conntrack_helper_in_ops);
 896  cleanup_inandlocalops:
 897         nf_unregister_hook(&ip_conntrack_local_out_ops);
 898  cleanup_inops:
 899         nf_unregister_hook(&ip_conntrack_in_ops);
 900  cleanup_defraglocalops:
 901         nf_unregister_hook(&ip_conntrack_defrag_local_out_ops);
 902  cleanup_defragops:
 903         nf_unregister_hook(&ip_conntrack_defrag_ops);
 904  cleanup_proc_stat:
 905 #ifdef CONFIG_PROC_FS
 906         remove_proc_entry("ip_conntrack", proc_net_stat);
 907  cleanup_proc_exp:
 908         proc_net_remove("ip_conntrack_expect");
 909  cleanup_proc:
 910         proc_net_remove("ip_conntrack");
 911  cleanup_init:
 912 #endif /* CONFIG_PROC_FS */
 913         ip_conntrack_cleanup();
 914  cleanup_nothing:
 915         return ret;
 916 }
 917
 918 /* FIXME: Allow NULL functions and sub in pointers to generic for
 919    them. --RR */
 920 int ip_conntrack_protocol_register(struct ip_conntrack_protocol *proto)
 921 {
 922         int ret = 0;
 923
 924         WRITE_LOCK(&ip_conntrack_lock);
 925         if (ip_ct_protos[proto->proto] != &ip_conntrack_generic_protocol) {
 926                 ret = -EBUSY;
 927                 goto out;
 928         }
 929         ip_ct_protos[proto->proto] = proto;
 930  out:
 931         WRITE_UNLOCK(&ip_conntrack_lock);
 932         return ret;
 933 }
 934
 935 void ip_conntrack_protocol_unregister(struct ip_conntrack_protocol *proto)
 936 {
 937         WRITE_LOCK(&ip_conntrack_lock);
 938         ip_ct_protos[proto->proto] = &ip_conntrack_generic_protocol;
 939         WRITE_UNLOCK(&ip_conntrack_lock);
 940
 941         /* Somebody could be still looking at the proto in bh. */
 942         synchronize_net();
 943
 944         /* Remove all contrack entries for this protocol */
 945         ip_ct_iterate_cleanup(kill_proto, &proto->proto);
 946 }
 947
 948 static int __init init(void)
 949 {
 950         return init_or_cleanup(1);
 951 }
 952
 953 static void __exit fini(void)
 954 {
 955         init_or_cleanup(0);
 956 }
 957
 958 module_init(init);
 959 module_exit(fini);
 960
 961 /* Some modules need us, but don't depend directly on any symbol.
 962    They should call this. */
 963 void need_ip_conntrack(void)
 964 {
 965 }
 966
 967 EXPORT_SYMBOL(ip_conntrack_protocol_register);
 968 EXPORT_SYMBOL(ip_conntrack_protocol_unregister);
 969 EXPORT_SYMBOL(ip_ct_get_tuple);
 970 EXPORT_SYMBOL(invert_tuplepr);
 971 EXPORT_SYMBOL(ip_conntrack_alter_reply);
 972 EXPORT_SYMBOL(ip_conntrack_destroyed);
 973 EXPORT_SYMBOL(need_ip_conntrack);
 974 EXPORT_SYMBOL(ip_conntrack_helper_register);
 975 EXPORT_SYMBOL(ip_conntrack_helper_unregister);
 976 EXPORT_SYMBOL(ip_ct_iterate_cleanup);
 977 EXPORT_SYMBOL(ip_ct_refresh_acct);
 978 EXPORT_SYMBOL(ip_ct_protos);
 979 EXPORT_SYMBOL(ip_ct_find_proto);
 980 EXPORT_SYMBOL(ip_conntrack_expect_alloc);
 981 EXPORT_SYMBOL(ip_conntrack_expect_free);
 982 EXPORT_SYMBOL(ip_conntrack_expect_related);
 983 EXPORT_SYMBOL(ip_conntrack_unexpect_related);
 984 EXPORT_SYMBOL(ip_conntrack_tuple_taken);
 985 EXPORT_SYMBOL(ip_ct_gather_frags);
 986 EXPORT_SYMBOL(ip_conntrack_htable_size);
 987 EXPORT_SYMBOL(ip_conntrack_lock);
 988 EXPORT_SYMBOL(ip_conntrack_hash);
 989 EXPORT_SYMBOL(ip_conntrack_untracked);
 990 EXPORT_SYMBOL_GPL(ip_conntrack_find_get);
 991 EXPORT_SYMBOL_GPL(ip_conntrack_put);
 992 #ifdef CONFIG_IP_NF_NAT_NEEDED
 993 EXPORT_SYMBOL(ip_conntrack_tcp_update);
 994 #endif