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         e = e->next;
 260
 261         if (e == &ip_conntrack_expect_list)
 262                 return NULL;
 263
 264         return e;
 265 }
 266
 267 static void exp_seq_stop(struct seq_file *s, void *v)
 268 {
 269         READ_UNLOCK(&ip_conntrack_lock);
 270 }
 271
 272 static int exp_seq_show(struct seq_file *s, void *v)
 273 {
 274         struct ip_conntrack_expect *expect = v;
 275
 276         if (expect->timeout.function)
 277                 seq_printf(s, "%ld ", timer_pending(&expect->timeout)
 278                            ? (long)(expect->timeout.expires - jiffies)/HZ : 0);
 279         else
 280                 seq_printf(s, "- ");
 281
 282         seq_printf(s, "proto=%u ", expect->tuple.dst.protonum);
 283
 284         print_tuple(s, &expect->tuple,
 285                     ip_ct_find_proto(expect->tuple.dst.protonum));
 286         return seq_putc(s, '\n');
 287 }
 288
 289 static struct seq_operations exp_seq_ops = {
 290         .start = exp_seq_start,
 291         .next = exp_seq_next,
 292         .stop = exp_seq_stop,
 293         .show = exp_seq_show
 294 };
 295
 296 static int exp_open(struct inode *inode, struct file *file)
 297 {
 298         return seq_open(file, &exp_seq_ops);
 299 }
 300
 301 static struct file_operations exp_file_ops = {
 302         .owner   = THIS_MODULE,
 303         .open    = exp_open,
 304         .read    = seq_read,
 305         .llseek  = seq_lseek,
 306         .release = seq_release
 307 };
 308
 309 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 310 {
 311         int cpu;
 312
 313         if (*pos == 0)
 314                 return SEQ_START_TOKEN;
 315
 316         for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
 317                 if (!cpu_possible(cpu))
 318                         continue;
 319                 *pos = cpu+1;
 320                 return &per_cpu(ip_conntrack_stat, cpu);
 321         }
 322
 323         return NULL;
 324 }
 325
 326 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 327 {
 328         int cpu;
 329
 330         for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
 331                 if (!cpu_possible(cpu))
 332                         continue;
 333                 *pos = cpu+1;
 334                 return &per_cpu(ip_conntrack_stat, cpu);
 335         }
 336
 337         return NULL;
 338 }
 339
 340 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 341 {
 342 }
 343
 344 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 345 {
 346         unsigned int nr_conntracks = atomic_read(&ip_conntrack_count);
 347         struct ip_conntrack_stat *st = v;
 348
 349         if (v == SEQ_START_TOKEN) {
 350                 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");
 351                 return 0;
 352         }
 353
 354         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 355                         "%08x %08x %08x %08x %08x  %08x %08x %08x \n",
 356                    nr_conntracks,
 357                    st->searched,
 358                    st->found,
 359                    st->new,
 360                    st->invalid,
 361                    st->ignore,
 362                    st->delete,
 363                    st->delete_list,
 364                    st->insert,
 365                    st->insert_failed,
 366                    st->drop,
 367                    st->early_drop,
 368                    st->error,
 369
 370                    st->expect_new,
 371                    st->expect_create,
 372                    st->expect_delete
 373                 );
 374         return 0;
 375 }
 376
 377 static struct seq_operations ct_cpu_seq_ops = {
 378         .start  = ct_cpu_seq_start,
 379         .next   = ct_cpu_seq_next,
 380         .stop   = ct_cpu_seq_stop,
 381         .show   = ct_cpu_seq_show,
 382 };
 383
 384 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 385 {
 386         return seq_open(file, &ct_cpu_seq_ops);
 387 }
 388
 389 static struct file_operations ct_cpu_seq_fops = {
 390         .owner   = THIS_MODULE,
 391         .open    = ct_cpu_seq_open,
 392         .read    = seq_read,
 393         .llseek  = seq_lseek,
 394         .release = seq_release_private,
 395 };
 396 #endif
 397
 398 static unsigned int ip_confirm(unsigned int hooknum,
 399                                struct sk_buff **pskb,
 400                                const struct net_device *in,
 401                                const struct net_device *out,
 402                                int (*okfn)(struct sk_buff *))
 403 {
 404         struct ip_conntrack *ct;
 405         enum ip_conntrack_info ctinfo;
 406
 407         /* This is where we call the helper: as the packet goes out. */
 408         ct = ip_conntrack_get(*pskb, &ctinfo);
 409         if (ct && ct->helper) {
 410                 unsigned int ret;
 411                 ret = ct->helper->help(pskb, ct, ctinfo);
 412                 if (ret != NF_ACCEPT)
 413                         return ret;
 414         }
 415
 416         /* We've seen it coming out the other side: confirm it */
 417         return ip_conntrack_confirm(pskb);
 418 }
 419
 420 static unsigned int ip_conntrack_defrag(unsigned int hooknum,
 421                                         struct sk_buff **pskb,
 422                                         const struct net_device *in,
 423                                         const struct net_device *out,
 424                                         int (*okfn)(struct sk_buff *))
 425 {
 426 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
 427         /* Previously seen (loopback)?  Ignore.  Do this before
 428            fragment check. */
 429         if ((*pskb)->nfct)
 430                 return NF_ACCEPT;
 431 #endif
 432
 433         /* Gather fragments. */
 434         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 435                 *pskb = ip_ct_gather_frags(*pskb,
 436                                            hooknum == NF_IP_PRE_ROUTING ?
 437                                            IP_DEFRAG_CONNTRACK_IN :
 438                                            IP_DEFRAG_CONNTRACK_OUT);
 439                 if (!*pskb)
 440                         return NF_STOLEN;
 441         }
 442         return NF_ACCEPT;
 443 }
 444
 445 static unsigned int ip_refrag(unsigned int hooknum,
 446                               struct sk_buff **pskb,
 447                               const struct net_device *in,
 448                               const struct net_device *out,
 449                               int (*okfn)(struct sk_buff *))
 450 {
 451         struct rtable *rt = (struct rtable *)(*pskb)->dst;
 452
 453         /* We've seen it coming out the other side: confirm */
 454         if (ip_confirm(hooknum, pskb, in, out, okfn) != NF_ACCEPT)
 455                 return NF_DROP;
 456
 457         /* Local packets are never produced too large for their
 458            interface.  We degfragment them at LOCAL_OUT, however,
 459            so we have to refragment them here. */
 460         if ((*pskb)->len > dst_mtu(&rt->u.dst) &&
 461             !skb_shinfo(*pskb)->tso_size) {
 462                 /* No hook can be after us, so this should be OK. */
 463                 ip_fragment(*pskb, okfn);
 464                 return NF_STOLEN;
 465         }
 466         return NF_ACCEPT;
 467 }
 468
 469 static unsigned int ip_conntrack_local(unsigned int hooknum,
 470                                        struct sk_buff **pskb,
 471                                        const struct net_device *in,
 472                                        const struct net_device *out,
 473                                        int (*okfn)(struct sk_buff *))
 474 {
 475         /* root is playing with raw sockets. */
 476         if ((*pskb)->len < sizeof(struct iphdr)
 477             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 478                 if (net_ratelimit())
 479                         printk("ipt_hook: happy cracking.\n");
 480                 return NF_ACCEPT;
 481         }
 482         return ip_conntrack_in(hooknum, pskb, in, out, okfn);
 483 }
 484
 485 /* Connection tracking may drop packets, but never alters them, so
 486    make it the first hook. */
 487 static struct nf_hook_ops ip_conntrack_defrag_ops = {
 488         .hook           = ip_conntrack_defrag,
 489         .owner          = THIS_MODULE,
 490         .pf             = PF_INET,
 491         .hooknum        = NF_IP_PRE_ROUTING,
 492         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 493 };
 494
 495 static struct nf_hook_ops ip_conntrack_in_ops = {
 496         .hook           = ip_conntrack_in,
 497         .owner          = THIS_MODULE,
 498         .pf             = PF_INET,
 499         .hooknum        = NF_IP_PRE_ROUTING,
 500         .priority       = NF_IP_PRI_CONNTRACK,
 501 };
 502
 503 static struct nf_hook_ops ip_conntrack_defrag_local_out_ops = {
 504         .hook           = ip_conntrack_defrag,
 505         .owner          = THIS_MODULE,
 506         .pf             = PF_INET,
 507         .hooknum        = NF_IP_LOCAL_OUT,
 508         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 509 };
 510
 511 static struct nf_hook_ops ip_conntrack_local_out_ops = {
 512         .hook           = ip_conntrack_local,
 513         .owner          = THIS_MODULE,
 514         .pf             = PF_INET,
 515         .hooknum        = NF_IP_LOCAL_OUT,
 516         .priority       = NF_IP_PRI_CONNTRACK,
 517 };
 518
 519 /* Refragmenter; last chance. */
 520 static struct nf_hook_ops ip_conntrack_out_ops = {
 521         .hook           = ip_refrag,
 522         .owner          = THIS_MODULE,
 523         .pf             = PF_INET,
 524         .hooknum        = NF_IP_POST_ROUTING,
 525         .priority       = NF_IP_PRI_LAST,
 526 };
 527
 528 static struct nf_hook_ops ip_conntrack_local_in_ops = {
 529         .hook           = ip_confirm,
 530         .owner          = THIS_MODULE,
 531         .pf             = PF_INET,
 532         .hooknum        = NF_IP_LOCAL_IN,
 533         .priority       = NF_IP_PRI_LAST-1,
 534 };
 535
 536 /* Sysctl support */
 537
 538 #ifdef CONFIG_SYSCTL
 539
 540 /* From ip_conntrack_core.c */
 541 extern int ip_conntrack_max;
 542 extern unsigned int ip_conntrack_htable_size;
 543
 544 /* From ip_conntrack_proto_tcp.c */
 545 extern unsigned long ip_ct_tcp_timeout_syn_sent;
 546 extern unsigned long ip_ct_tcp_timeout_syn_recv;
 547 extern unsigned long ip_ct_tcp_timeout_established;
 548 extern unsigned long ip_ct_tcp_timeout_fin_wait;
 549 extern unsigned long ip_ct_tcp_timeout_close_wait;
 550 extern unsigned long ip_ct_tcp_timeout_last_ack;
 551 extern unsigned long ip_ct_tcp_timeout_time_wait;
 552 extern unsigned long ip_ct_tcp_timeout_close;
 553 extern unsigned long ip_ct_tcp_timeout_max_retrans;
 554 extern int ip_ct_tcp_loose;
 555 extern int ip_ct_tcp_be_liberal;
 556 extern int ip_ct_tcp_max_retrans;
 557
 558 /* From ip_conntrack_proto_udp.c */
 559 extern unsigned long ip_ct_udp_timeout;
 560 extern unsigned long ip_ct_udp_timeout_stream;
 561
 562 /* From ip_conntrack_proto_icmp.c */
 563 extern unsigned long ip_ct_icmp_timeout;
 564
 565 /* From ip_conntrack_proto_icmp.c */
 566 extern unsigned long ip_ct_generic_timeout;
 567
 568 /* Log invalid packets of a given protocol */
 569 static int log_invalid_proto_min = 0;
 570 static int log_invalid_proto_max = 255;
 571
 572 static struct ctl_table_header *ip_ct_sysctl_header;
 573
 574 static ctl_table ip_ct_sysctl_table[] = {
 575         {
 576                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 577                 .procname       = "ip_conntrack_max",
 578                 .data           = &ip_conntrack_max,
 579                 .maxlen         = sizeof(int),
 580                 .mode           = 0644,
 581                 .proc_handler   = &proc_dointvec,
 582         },
 583         {
 584                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 585                 .procname       = "ip_conntrack_count",
 586                 .data           = &ip_conntrack_count,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0444,
 589                 .proc_handler   = &proc_dointvec,
 590         },
 591         {
 592                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 593                 .procname       = "ip_conntrack_buckets",
 594                 .data           = &ip_conntrack_htable_size,
 595                 .maxlen         = sizeof(unsigned int),
 596                 .mode           = 0444,
 597                 .proc_handler   = &proc_dointvec,
 598         },
 599         {
 600                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_SENT,
 601                 .procname       = "ip_conntrack_tcp_timeout_syn_sent",
 602                 .data           = &ip_ct_tcp_timeout_syn_sent,
 603                 .maxlen         = sizeof(unsigned int),
 604                 .mode           = 0644,
 605                 .proc_handler   = &proc_dointvec_jiffies,
 606         },
 607         {
 608                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_RECV,
 609                 .procname       = "ip_conntrack_tcp_timeout_syn_recv",
 610                 .data           = &ip_ct_tcp_timeout_syn_recv,
 611                 .maxlen         = sizeof(unsigned int),
 612                 .mode           = 0644,
 613                 .proc_handler   = &proc_dointvec_jiffies,
 614         },
 615         {
 616                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_ESTABLISHED,
 617                 .procname       = "ip_conntrack_tcp_timeout_established",
 618                 .data           = &ip_ct_tcp_timeout_established,
 619                 .maxlen         = sizeof(unsigned int),
 620                 .mode           = 0644,
 621                 .proc_handler   = &proc_dointvec_jiffies,
 622         },
 623         {
 624                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_FIN_WAIT,
 625                 .procname       = "ip_conntrack_tcp_timeout_fin_wait",
 626                 .data           = &ip_ct_tcp_timeout_fin_wait,
 627                 .maxlen         = sizeof(unsigned int),
 628                 .mode           = 0644,
 629                 .proc_handler   = &proc_dointvec_jiffies,
 630         },
 631         {
 632                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE_WAIT,
 633                 .procname       = "ip_conntrack_tcp_timeout_close_wait",
 634                 .data           = &ip_ct_tcp_timeout_close_wait,
 635                 .maxlen         = sizeof(unsigned int),
 636                 .mode           = 0644,
 637                 .proc_handler   = &proc_dointvec_jiffies,
 638         },
 639         {
 640                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_LAST_ACK,
 641                 .procname       = "ip_conntrack_tcp_timeout_last_ack",
 642                 .data           = &ip_ct_tcp_timeout_last_ack,
 643                 .maxlen         = sizeof(unsigned int),
 644                 .mode           = 0644,
 645                 .proc_handler   = &proc_dointvec_jiffies,
 646         },
 647         {
 648                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_TIME_WAIT,
 649                 .procname       = "ip_conntrack_tcp_timeout_time_wait",
 650                 .data           = &ip_ct_tcp_timeout_time_wait,
 651                 .maxlen         = sizeof(unsigned int),
 652                 .mode           = 0644,
 653                 .proc_handler   = &proc_dointvec_jiffies,
 654         },
 655         {
 656                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE,
 657                 .procname       = "ip_conntrack_tcp_timeout_close",
 658                 .data           = &ip_ct_tcp_timeout_close,
 659                 .maxlen         = sizeof(unsigned int),
 660                 .mode           = 0644,
 661                 .proc_handler   = &proc_dointvec_jiffies,
 662         },
 663         {
 664                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT,
 665                 .procname       = "ip_conntrack_udp_timeout",
 666                 .data           = &ip_ct_udp_timeout,
 667                 .maxlen         = sizeof(unsigned int),
 668                 .mode           = 0644,
 669                 .proc_handler   = &proc_dointvec_jiffies,
 670         },
 671         {
 672                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT_STREAM,
 673                 .procname       = "ip_conntrack_udp_timeout_stream",
 674                 .data           = &ip_ct_udp_timeout_stream,
 675                 .maxlen         = sizeof(unsigned int),
 676                 .mode           = 0644,
 677                 .proc_handler   = &proc_dointvec_jiffies,
 678         },
 679         {
 680                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 681                 .procname       = "ip_conntrack_icmp_timeout",
 682                 .data           = &ip_ct_icmp_timeout,
 683                 .maxlen         = sizeof(unsigned int),
 684                 .mode           = 0644,
 685                 .proc_handler   = &proc_dointvec_jiffies,
 686         },
 687         {
 688                 .ctl_name       = NET_IPV4_NF_CONNTRACK_GENERIC_TIMEOUT,
 689                 .procname       = "ip_conntrack_generic_timeout",
 690                 .data           = &ip_ct_generic_timeout,
 691                 .maxlen         = sizeof(unsigned int),
 692                 .mode           = 0644,
 693                 .proc_handler   = &proc_dointvec_jiffies,
 694         },
 695         {
 696                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 697                 .procname       = "ip_conntrack_log_invalid",
 698                 .data           = &ip_ct_log_invalid,
 699                 .maxlen         = sizeof(unsigned int),
 700                 .mode           = 0644,
 701                 .proc_handler   = &proc_dointvec_minmax,
 702                 .strategy       = &sysctl_intvec,
 703                 .extra1         = &log_invalid_proto_min,
 704                 .extra2         = &log_invalid_proto_max,
 705         },
 706         {
 707                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_MAX_RETRANS,
 708                 .procname       = "ip_conntrack_tcp_timeout_max_retrans",
 709                 .data           = &ip_ct_tcp_timeout_max_retrans,
 710                 .maxlen         = sizeof(unsigned int),
 711                 .mode           = 0644,
 712                 .proc_handler   = &proc_dointvec_jiffies,
 713         },
 714         {
 715                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_LOOSE,
 716                 .procname       = "ip_conntrack_tcp_loose",
 717                 .data           = &ip_ct_tcp_loose,
 718                 .maxlen         = sizeof(unsigned int),
 719                 .mode           = 0644,
 720                 .proc_handler   = &proc_dointvec,
 721         },
 722         {
 723                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_BE_LIBERAL,
 724                 .procname       = "ip_conntrack_tcp_be_liberal",
 725                 .data           = &ip_ct_tcp_be_liberal,
 726                 .maxlen         = sizeof(unsigned int),
 727                 .mode           = 0644,
 728                 .proc_handler   = &proc_dointvec,
 729         },
 730         {
 731                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_MAX_RETRANS,
 732                 .procname       = "ip_conntrack_tcp_max_retrans",
 733                 .data           = &ip_ct_tcp_max_retrans,
 734                 .maxlen         = sizeof(unsigned int),
 735                 .mode           = 0644,
 736                 .proc_handler   = &proc_dointvec,
 737         },
 738         { .ctl_name = 0 }
 739 };
 740
 741 #define NET_IP_CONNTRACK_MAX 2089
 742
 743 static ctl_table ip_ct_netfilter_table[] = {
 744         {
 745                 .ctl_name       = NET_IPV4_NETFILTER,
 746                 .procname       = "netfilter",
 747                 .mode           = 0555,
 748                 .child          = ip_ct_sysctl_table,
 749         },
 750         {
 751                 .ctl_name       = NET_IP_CONNTRACK_MAX,
 752                 .procname       = "ip_conntrack_max",
 753                 .data           = &ip_conntrack_max,
 754                 .maxlen         = sizeof(int),
 755                 .mode           = 0644,
 756                 .proc_handler   = &proc_dointvec
 757         },
 758         { .ctl_name = 0 }
 759 };
 760
 761 static ctl_table ip_ct_ipv4_table[] = {
 762         {
 763                 .ctl_name       = NET_IPV4,
 764                 .procname       = "ipv4",
 765                 .mode           = 0555,
 766                 .child          = ip_ct_netfilter_table,
 767         },
 768         { .ctl_name = 0 }
 769 };
 770
 771 static ctl_table ip_ct_net_table[] = {
 772         {
 773                 .ctl_name       = CTL_NET,
 774                 .procname       = "net",
 775                 .mode           = 0555,
 776                 .child          = ip_ct_ipv4_table,
 777         },
 778         { .ctl_name = 0 }
 779 };
 780
 781 EXPORT_SYMBOL(ip_ct_log_invalid);
 782 #endif /* CONFIG_SYSCTL */
 783
 784 static int init_or_cleanup(int init)
 785 {
 786 #ifdef CONFIG_PROC_FS
 787         struct proc_dir_entry *proc, *proc_exp, *proc_stat;
 788 #endif
 789         int ret = 0;
 790
 791         if (!init) goto cleanup;
 792
 793         ret = ip_conntrack_init();
 794         if (ret < 0)
 795                 goto cleanup_nothing;
 796
 797 #ifdef CONFIG_PROC_FS
 798         ret = -ENOMEM;
 799         proc = proc_net_fops_create("ip_conntrack", 0440, &ct_file_ops);
 800         if (!proc) goto cleanup_init;
 801
 802         proc_exp = proc_net_fops_create("ip_conntrack_expect", 0440,
 803                                         &exp_file_ops);
 804         if (!proc_exp) goto cleanup_proc;
 805
 806         proc_stat = create_proc_entry("ip_conntrack", S_IRUGO, proc_net_stat);
 807         if (!proc_stat)
 808                 goto cleanup_proc_exp;
 809
 810         proc_stat->proc_fops = &ct_cpu_seq_fops;
 811         proc_stat->owner = THIS_MODULE;
 812 #endif
 813
 814         ret = nf_register_hook(&ip_conntrack_defrag_ops);
 815         if (ret < 0) {
 816                 printk("ip_conntrack: can't register pre-routing defrag hook.\n");
 817                 goto cleanup_proc_stat;
 818         }
 819         ret = nf_register_hook(&ip_conntrack_defrag_local_out_ops);
 820         if (ret < 0) {
 821                 printk("ip_conntrack: can't register local_out defrag hook.\n");
 822                 goto cleanup_defragops;
 823         }
 824         ret = nf_register_hook(&ip_conntrack_in_ops);
 825         if (ret < 0) {
 826                 printk("ip_conntrack: can't register pre-routing hook.\n");
 827                 goto cleanup_defraglocalops;
 828         }
 829         ret = nf_register_hook(&ip_conntrack_local_out_ops);
 830         if (ret < 0) {
 831                 printk("ip_conntrack: can't register local out hook.\n");
 832                 goto cleanup_inops;
 833         }
 834         ret = nf_register_hook(&ip_conntrack_out_ops);
 835         if (ret < 0) {
 836                 printk("ip_conntrack: can't register post-routing hook.\n");
 837                 goto cleanup_inandlocalops;
 838         }
 839         ret = nf_register_hook(&ip_conntrack_local_in_ops);
 840         if (ret < 0) {
 841                 printk("ip_conntrack: can't register local in hook.\n");
 842                 goto cleanup_inoutandlocalops;
 843         }
 844 #ifdef CONFIG_SYSCTL
 845         ip_ct_sysctl_header = register_sysctl_table(ip_ct_net_table, 0);
 846         if (ip_ct_sysctl_header == NULL) {
 847                 printk("ip_conntrack: can't register to sysctl.\n");
 848                 ret = -ENOMEM;
 849                 goto cleanup_localinops;
 850         }
 851 #endif
 852
 853         return ret;
 854
 855  cleanup:
 856 #ifdef CONFIG_SYSCTL
 857         unregister_sysctl_table(ip_ct_sysctl_header);
 858  cleanup_localinops:
 859 #endif
 860         nf_unregister_hook(&ip_conntrack_local_in_ops);
 861  cleanup_inoutandlocalops:
 862         nf_unregister_hook(&ip_conntrack_out_ops);
 863  cleanup_inandlocalops:
 864         nf_unregister_hook(&ip_conntrack_local_out_ops);
 865  cleanup_inops:
 866         nf_unregister_hook(&ip_conntrack_in_ops);
 867  cleanup_defraglocalops:
 868         nf_unregister_hook(&ip_conntrack_defrag_local_out_ops);
 869  cleanup_defragops:
 870         nf_unregister_hook(&ip_conntrack_defrag_ops);
 871  cleanup_proc_stat:
 872 #ifdef CONFIG_PROC_FS
 873         remove_proc_entry("ip_conntrack", proc_net_stat);
 874  cleanup_proc_exp:
 875         proc_net_remove("ip_conntrack_expect");
 876  cleanup_proc:
 877         proc_net_remove("ip_conntrack");
 878  cleanup_init:
 879 #endif /* CONFIG_PROC_FS */
 880         ip_conntrack_cleanup();
 881  cleanup_nothing:
 882         return ret;
 883 }
 884
 885 /* FIXME: Allow NULL functions and sub in pointers to generic for
 886    them. --RR */
 887 int ip_conntrack_protocol_register(struct ip_conntrack_protocol *proto)
 888 {
 889         int ret = 0;
 890
 891         WRITE_LOCK(&ip_conntrack_lock);
 892         if (ip_ct_protos[proto->proto] != &ip_conntrack_generic_protocol) {
 893                 ret = -EBUSY;
 894                 goto out;
 895         }
 896         ip_ct_protos[proto->proto] = proto;
 897  out:
 898         WRITE_UNLOCK(&ip_conntrack_lock);
 899         return ret;
 900 }
 901
 902 void ip_conntrack_protocol_unregister(struct ip_conntrack_protocol *proto)
 903 {
 904         WRITE_LOCK(&ip_conntrack_lock);
 905         ip_ct_protos[proto->proto] = &ip_conntrack_generic_protocol;
 906         WRITE_UNLOCK(&ip_conntrack_lock);
 907
 908         /* Somebody could be still looking at the proto in bh. */
 909         synchronize_net();
 910
 911         /* Remove all contrack entries for this protocol */
 912         ip_ct_iterate_cleanup(kill_proto, &proto->proto);
 913 }
 914
 915 static int __init init(void)
 916 {
 917         return init_or_cleanup(1);
 918 }
 919
 920 static void __exit fini(void)
 921 {
 922         init_or_cleanup(0);
 923 }
 924
 925 module_init(init);
 926 module_exit(fini);
 927
 928 /* Some modules need us, but don't depend directly on any symbol.
 929    They should call this. */
 930 void need_ip_conntrack(void)
 931 {
 932 }
 933
 934 EXPORT_SYMBOL(ip_conntrack_protocol_register);
 935 EXPORT_SYMBOL(ip_conntrack_protocol_unregister);
 936 EXPORT_SYMBOL(ip_ct_get_tuple);
 937 EXPORT_SYMBOL(invert_tuplepr);
 938 EXPORT_SYMBOL(ip_conntrack_alter_reply);
 939 EXPORT_SYMBOL(ip_conntrack_destroyed);
 940 EXPORT_SYMBOL(need_ip_conntrack);
 941 EXPORT_SYMBOL(ip_conntrack_helper_register);
 942 EXPORT_SYMBOL(ip_conntrack_helper_unregister);
 943 EXPORT_SYMBOL(ip_ct_iterate_cleanup);
 944 EXPORT_SYMBOL(ip_ct_refresh_acct);
 945 EXPORT_SYMBOL(ip_ct_protos);
 946 EXPORT_SYMBOL(ip_ct_find_proto);
 947 EXPORT_SYMBOL(ip_conntrack_expect_alloc);
 948 EXPORT_SYMBOL(ip_conntrack_expect_free);
 949 EXPORT_SYMBOL(ip_conntrack_expect_related);
 950 EXPORT_SYMBOL(ip_conntrack_unexpect_related);
 951 EXPORT_SYMBOL(ip_conntrack_tuple_taken);
 952 EXPORT_SYMBOL(ip_ct_gather_frags);
 953 EXPORT_SYMBOL(ip_conntrack_htable_size);
 954 EXPORT_SYMBOL(ip_conntrack_lock);
 955 EXPORT_SYMBOL(ip_conntrack_hash);
 956 EXPORT_SYMBOL(ip_conntrack_untracked);
 957 EXPORT_SYMBOL_GPL(ip_conntrack_find_get);
 958 EXPORT_SYMBOL_GPL(ip_conntrack_put);
 959 #ifdef CONFIG_IP_NF_NAT_NEEDED
 960 EXPORT_SYMBOL(ip_conntrack_tcp_update);
 961 #endif