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 defined(CONFIG_VNET) || defined(CONFIG_VNET_MODULE)
 170         if (seq_printf(s, "xid=%d\n", conntrack->xid[IP_CT_DIR_ORIGINAL]))
 171                 return 1;
 172 #endif
 173
 174         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_ORIGINAL]))
 175                 return -ENOSPC;
 176
 177         if (!(test_bit(IPS_SEEN_REPLY_BIT, &conntrack->status)))
 178                 if (seq_printf(s, "[UNREPLIED] "))
 179                         return -ENOSPC;
 180
 181         if (print_tuple(s, &conntrack->tuplehash[IP_CT_DIR_REPLY].tuple,
 182                         proto))
 183                 return -ENOSPC;
 184
 185 #if defined(CONFIG_VNET) || defined(CONFIG_VNET_MODULE)
 186         if (seq_printf(s, "xid=%d\n", conntrack->xid[IP_CT_DIR_REPLY]))
 187                 return 1;
 188 #endif
 189
 190         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_REPLY]))
 191                 return -ENOSPC;
 192
 193         if (test_bit(IPS_ASSURED_BIT, &conntrack->status))
 194                 if (seq_printf(s, "[ASSURED] "))
 195                         return -ENOSPC;
 196
 197 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 198         if (seq_printf(s, "mark=%lu ", conntrack->mark))
 199                 return -ENOSPC;
 200 #endif
 201
 202         if (seq_printf(s, "use=%u\n", atomic_read(&conntrack->ct_general.use)))
 203                 return -ENOSPC;
 204
 205         return 0;
 206 }
 207
 208 static struct seq_operations ct_seq_ops = {
 209         .start = ct_seq_start,
 210         .next  = ct_seq_next,
 211         .stop  = ct_seq_stop,
 212         .show  = ct_seq_show
 213 };
 214
 215 static int ct_open(struct inode *inode, struct file *file)
 216 {
 217         struct seq_file *seq;
 218         struct ct_iter_state *st;
 219         int ret;
 220
 221         st = kmalloc(sizeof(struct ct_iter_state), GFP_KERNEL);
 222         if (st == NULL)
 223                 return -ENOMEM;
 224         ret = seq_open(file, &ct_seq_ops);
 225         if (ret)
 226                 goto out_free;
 227         seq          = file->private_data;
 228         seq->private = st;
 229         memset(st, 0, sizeof(struct ct_iter_state));
 230         return ret;
 231 out_free:
 232         kfree(st);
 233         return ret;
 234 }
 235
 236 static struct file_operations ct_file_ops = {
 237         .owner   = THIS_MODULE,
 238         .open    = ct_open,
 239         .read    = seq_read,
 240         .llseek  = seq_lseek,
 241         .release = seq_release_private,
 242 };
 243
 244 /* expects */
 245 static void *exp_seq_start(struct seq_file *s, loff_t *pos)
 246 {
 247         struct list_head *e = &ip_conntrack_expect_list;
 248         loff_t i;
 249
 250         /* strange seq_file api calls stop even if we fail,
 251          * thus we need to grab lock since stop unlocks */
 252         READ_LOCK(&ip_conntrack_lock);
 253
 254         if (list_empty(e))
 255                 return NULL;
 256
 257         for (i = 0; i <= *pos; i++) {
 258                 e = e->next;
 259                 if (e == &ip_conntrack_expect_list)
 260                         return NULL;
 261         }
 262         return e;
 263 }
 264
 265 static void *exp_seq_next(struct seq_file *s, void *v, loff_t *pos)
 266 {
 267         struct list_head *e = v;
 268
 269         ++*pos;
 270         e = e->next;
 271
 272         if (e == &ip_conntrack_expect_list)
 273                 return NULL;
 274
 275         return e;
 276 }
 277
 278 static void exp_seq_stop(struct seq_file *s, void *v)
 279 {
 280         READ_UNLOCK(&ip_conntrack_lock);
 281 }
 282
 283 static int exp_seq_show(struct seq_file *s, void *v)
 284 {
 285         struct ip_conntrack_expect *expect = v;
 286
 287         if (expect->timeout.function)
 288                 seq_printf(s, "%ld ", timer_pending(&expect->timeout)
 289                            ? (long)(expect->timeout.expires - jiffies)/HZ : 0);
 290         else
 291                 seq_printf(s, "- ");
 292
 293         seq_printf(s, "proto=%u ", expect->tuple.dst.protonum);
 294
 295         print_tuple(s, &expect->tuple,
 296                     ip_ct_find_proto(expect->tuple.dst.protonum));
 297         return seq_putc(s, '\n');
 298 }
 299
 300 static struct seq_operations exp_seq_ops = {
 301         .start = exp_seq_start,
 302         .next = exp_seq_next,
 303         .stop = exp_seq_stop,
 304         .show = exp_seq_show
 305 };
 306
 307 static int exp_open(struct inode *inode, struct file *file)
 308 {
 309         return seq_open(file, &exp_seq_ops);
 310 }
 311
 312 static struct file_operations exp_file_ops = {
 313         .owner   = THIS_MODULE,
 314         .open    = exp_open,
 315         .read    = seq_read,
 316         .llseek  = seq_lseek,
 317         .release = seq_release
 318 };
 319
 320 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 321 {
 322         int cpu;
 323
 324         if (*pos == 0)
 325                 return SEQ_START_TOKEN;
 326
 327         for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
 328                 if (!cpu_possible(cpu))
 329                         continue;
 330                 *pos = cpu+1;
 331                 return &per_cpu(ip_conntrack_stat, cpu);
 332         }
 333
 334         return NULL;
 335 }
 336
 337 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 338 {
 339         int cpu;
 340
 341         for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
 342                 if (!cpu_possible(cpu))
 343                         continue;
 344                 *pos = cpu+1;
 345                 return &per_cpu(ip_conntrack_stat, cpu);
 346         }
 347
 348         return NULL;
 349 }
 350
 351 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 352 {
 353 }
 354
 355 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 356 {
 357         unsigned int nr_conntracks = atomic_read(&ip_conntrack_count);
 358         struct ip_conntrack_stat *st = v;
 359
 360         if (v == SEQ_START_TOKEN) {
 361                 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");
 362                 return 0;
 363         }
 364
 365         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 366                         "%08x %08x %08x %08x %08x  %08x %08x %08x \n",
 367                    nr_conntracks,
 368                    st->searched,
 369                    st->found,
 370                    st->new,
 371                    st->invalid,
 372                    st->ignore,
 373                    st->delete,
 374                    st->delete_list,
 375                    st->insert,
 376                    st->insert_failed,
 377                    st->drop,
 378                    st->early_drop,
 379                    st->error,
 380
 381                    st->expect_new,
 382                    st->expect_create,
 383                    st->expect_delete
 384                 );
 385         return 0;
 386 }
 387
 388 static struct seq_operations ct_cpu_seq_ops = {
 389         .start  = ct_cpu_seq_start,
 390         .next   = ct_cpu_seq_next,
 391         .stop   = ct_cpu_seq_stop,
 392         .show   = ct_cpu_seq_show,
 393 };
 394
 395 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 396 {
 397         return seq_open(file, &ct_cpu_seq_ops);
 398 }
 399
 400 static struct file_operations ct_cpu_seq_fops = {
 401         .owner   = THIS_MODULE,
 402         .open    = ct_cpu_seq_open,
 403         .read    = seq_read,
 404         .llseek  = seq_lseek,
 405         .release = seq_release_private,
 406 };
 407 #endif
 408
 409 static unsigned int ip_confirm(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         /* We've seen it coming out the other side: confirm it */
 416         return ip_conntrack_confirm(pskb);
 417 }
 418
 419 static unsigned int ip_conntrack_help(unsigned int hooknum,
 420                                       struct sk_buff **pskb,
 421                                       const struct net_device *in,
 422                                       const struct net_device *out,
 423                                       int (*okfn)(struct sk_buff *))
 424 {
 425         struct ip_conntrack *ct;
 426         enum ip_conntrack_info ctinfo;
 427
 428         /* This is where we call the helper: as the packet goes out. */
 429         ct = ip_conntrack_get(*pskb, &ctinfo);
 430         if (ct && ct->helper) {
 431                 unsigned int ret;
 432                 ret = ct->helper->help(pskb, ct, ctinfo);
 433                 if (ret != NF_ACCEPT)
 434                         return ret;
 435         }
 436         return NF_ACCEPT;
 437 }
 438
 439 static unsigned int ip_conntrack_defrag(unsigned int hooknum,
 440                                         struct sk_buff **pskb,
 441                                         const struct net_device *in,
 442                                         const struct net_device *out,
 443                                         int (*okfn)(struct sk_buff *))
 444 {
 445 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
 446         /* Previously seen (loopback)?  Ignore.  Do this before
 447            fragment check. */
 448         if ((*pskb)->nfct)
 449                 return NF_ACCEPT;
 450 #endif
 451
 452         /* Gather fragments. */
 453         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 454                 *pskb = ip_ct_gather_frags(*pskb,
 455                                            hooknum == NF_IP_PRE_ROUTING ?
 456                                            IP_DEFRAG_CONNTRACK_IN :
 457                                            IP_DEFRAG_CONNTRACK_OUT);
 458                 if (!*pskb)
 459                         return NF_STOLEN;
 460         }
 461         return NF_ACCEPT;
 462 }
 463
 464 static unsigned int ip_refrag(unsigned int hooknum,
 465                               struct sk_buff **pskb,
 466                               const struct net_device *in,
 467                               const struct net_device *out,
 468                               int (*okfn)(struct sk_buff *))
 469 {
 470         struct rtable *rt = (struct rtable *)(*pskb)->dst;
 471
 472         /* We've seen it coming out the other side: confirm */
 473         if (ip_confirm(hooknum, pskb, in, out, okfn) != NF_ACCEPT)
 474                 return NF_DROP;
 475
 476         /* Local packets are never produced too large for their
 477            interface.  We degfragment them at LOCAL_OUT, however,
 478            so we have to refragment them here. */
 479         if ((*pskb)->len > dst_mtu(&rt->u.dst) &&
 480             !skb_shinfo(*pskb)->tso_size) {
 481                 /* No hook can be after us, so this should be OK. */
 482                 ip_fragment(*pskb, okfn);
 483                 return NF_STOLEN;
 484         }
 485         return NF_ACCEPT;
 486 }
 487
 488 static unsigned int ip_conntrack_local(unsigned int hooknum,
 489                                        struct sk_buff **pskb,
 490                                        const struct net_device *in,
 491                                        const struct net_device *out,
 492                                        int (*okfn)(struct sk_buff *))
 493 {
 494         /* root is playing with raw sockets. */
 495         if ((*pskb)->len < sizeof(struct iphdr)
 496             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 497                 if (net_ratelimit())
 498                         printk("ipt_hook: happy cracking.\n");
 499                 return NF_ACCEPT;
 500         }
 501         return ip_conntrack_in(hooknum, pskb, in, out, okfn);
 502 }
 503
 504 /* Connection tracking may drop packets, but never alters them, so
 505    make it the first hook. */
 506 static struct nf_hook_ops ip_conntrack_defrag_ops = {
 507         .hook           = ip_conntrack_defrag,
 508         .owner          = THIS_MODULE,
 509         .pf             = PF_INET,
 510         .hooknum        = NF_IP_PRE_ROUTING,
 511         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 512 };
 513
 514 static struct nf_hook_ops ip_conntrack_in_ops = {
 515         .hook           = ip_conntrack_in,
 516         .owner          = THIS_MODULE,
 517         .pf             = PF_INET,
 518         .hooknum        = NF_IP_PRE_ROUTING,
 519         .priority       = NF_IP_PRI_CONNTRACK,
 520 };
 521
 522 static struct nf_hook_ops ip_conntrack_defrag_local_out_ops = {
 523         .hook           = ip_conntrack_defrag,
 524         .owner          = THIS_MODULE,
 525         .pf             = PF_INET,
 526         .hooknum        = NF_IP_LOCAL_OUT,
 527         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 528 };
 529
 530 static struct nf_hook_ops ip_conntrack_local_out_ops = {
 531         .hook           = ip_conntrack_local,
 532         .owner          = THIS_MODULE,
 533         .pf             = PF_INET,
 534         .hooknum        = NF_IP_LOCAL_OUT,
 535         .priority       = NF_IP_PRI_CONNTRACK,
 536 };
 537
 538 /* helpers */
 539 static struct nf_hook_ops ip_conntrack_helper_out_ops = {
 540         .hook           = ip_conntrack_help,
 541         .owner          = THIS_MODULE,
 542         .pf             = PF_INET,
 543         .hooknum        = NF_IP_POST_ROUTING,
 544         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 545 };
 546
 547 static struct nf_hook_ops ip_conntrack_helper_in_ops = {
 548         .hook           = ip_conntrack_help,
 549         .owner          = THIS_MODULE,
 550         .pf             = PF_INET,
 551         .hooknum        = NF_IP_LOCAL_IN,
 552         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 553 };
 554
 555 /* Refragmenter; last chance. */
 556 static struct nf_hook_ops ip_conntrack_out_ops = {
 557         .hook           = ip_refrag,
 558         .owner          = THIS_MODULE,
 559         .pf             = PF_INET,
 560         .hooknum        = NF_IP_POST_ROUTING,
 561         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 562 };
 563
 564 static struct nf_hook_ops ip_conntrack_local_in_ops = {
 565         .hook           = ip_confirm,
 566         .owner          = THIS_MODULE,
 567         .pf             = PF_INET,
 568         .hooknum        = NF_IP_LOCAL_IN,
 569         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 570 };
 571
 572 /* Sysctl support */
 573
 574 #ifdef CONFIG_SYSCTL
 575
 576 /* From ip_conntrack_core.c */
 577 extern int ip_conntrack_max;
 578 extern unsigned int ip_conntrack_htable_size;
 579
 580 /* From ip_conntrack_proto_tcp.c */
 581 extern unsigned long ip_ct_tcp_timeout_syn_sent;
 582 extern unsigned long ip_ct_tcp_timeout_syn_recv;
 583 extern unsigned long ip_ct_tcp_timeout_established;
 584 extern unsigned long ip_ct_tcp_timeout_fin_wait;
 585 extern unsigned long ip_ct_tcp_timeout_close_wait;
 586 extern unsigned long ip_ct_tcp_timeout_last_ack;
 587 extern unsigned long ip_ct_tcp_timeout_time_wait;
 588 extern unsigned long ip_ct_tcp_timeout_close;
 589 extern unsigned long ip_ct_tcp_timeout_max_retrans;
 590 extern int ip_ct_tcp_loose;
 591 extern int ip_ct_tcp_be_liberal;
 592 extern int ip_ct_tcp_max_retrans;
 593
 594 /* From ip_conntrack_proto_udp.c */
 595 extern unsigned long ip_ct_udp_timeout;
 596 extern unsigned long ip_ct_udp_timeout_stream;
 597
 598 /* From ip_conntrack_proto_icmp.c */
 599 extern unsigned long ip_ct_icmp_timeout;
 600
 601 /* From ip_conntrack_proto_icmp.c */
 602 extern unsigned long ip_ct_generic_timeout;
 603
 604 /* Log invalid packets of a given protocol */
 605 static int log_invalid_proto_min = 0;
 606 static int log_invalid_proto_max = 255;
 607
 608 static struct ctl_table_header *ip_ct_sysctl_header;
 609
 610 static ctl_table ip_ct_sysctl_table[] = {
 611         {
 612                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 613                 .procname       = "ip_conntrack_max",
 614                 .data           = &ip_conntrack_max,
 615                 .maxlen         = sizeof(int),
 616                 .mode           = 0644,
 617                 .proc_handler   = &proc_dointvec,
 618         },
 619         {
 620                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 621                 .procname       = "ip_conntrack_count",
 622                 .data           = &ip_conntrack_count,
 623                 .maxlen         = sizeof(int),
 624                 .mode           = 0444,
 625                 .proc_handler   = &proc_dointvec,
 626         },
 627         {
 628                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 629                 .procname       = "ip_conntrack_buckets",
 630                 .data           = &ip_conntrack_htable_size,
 631                 .maxlen         = sizeof(unsigned int),
 632                 .mode           = 0444,
 633                 .proc_handler   = &proc_dointvec,
 634         },
 635         {
 636                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_SENT,
 637                 .procname       = "ip_conntrack_tcp_timeout_syn_sent",
 638                 .data           = &ip_ct_tcp_timeout_syn_sent,
 639                 .maxlen         = sizeof(unsigned int),
 640                 .mode           = 0644,
 641                 .proc_handler   = &proc_dointvec_jiffies,
 642         },
 643         {
 644                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_RECV,
 645                 .procname       = "ip_conntrack_tcp_timeout_syn_recv",
 646                 .data           = &ip_ct_tcp_timeout_syn_recv,
 647                 .maxlen         = sizeof(unsigned int),
 648                 .mode           = 0644,
 649                 .proc_handler   = &proc_dointvec_jiffies,
 650         },
 651         {
 652                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_ESTABLISHED,
 653                 .procname       = "ip_conntrack_tcp_timeout_established",
 654                 .data           = &ip_ct_tcp_timeout_established,
 655                 .maxlen         = sizeof(unsigned int),
 656                 .mode           = 0644,
 657                 .proc_handler   = &proc_dointvec_jiffies,
 658         },
 659         {
 660                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_FIN_WAIT,
 661                 .procname       = "ip_conntrack_tcp_timeout_fin_wait",
 662                 .data           = &ip_ct_tcp_timeout_fin_wait,
 663                 .maxlen         = sizeof(unsigned int),
 664                 .mode           = 0644,
 665                 .proc_handler   = &proc_dointvec_jiffies,
 666         },
 667         {
 668                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE_WAIT,
 669                 .procname       = "ip_conntrack_tcp_timeout_close_wait",
 670                 .data           = &ip_ct_tcp_timeout_close_wait,
 671                 .maxlen         = sizeof(unsigned int),
 672                 .mode           = 0644,
 673                 .proc_handler   = &proc_dointvec_jiffies,
 674         },
 675         {
 676                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_LAST_ACK,
 677                 .procname       = "ip_conntrack_tcp_timeout_last_ack",
 678                 .data           = &ip_ct_tcp_timeout_last_ack,
 679                 .maxlen         = sizeof(unsigned int),
 680                 .mode           = 0644,
 681                 .proc_handler   = &proc_dointvec_jiffies,
 682         },
 683         {
 684                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_TIME_WAIT,
 685                 .procname       = "ip_conntrack_tcp_timeout_time_wait",
 686                 .data           = &ip_ct_tcp_timeout_time_wait,
 687                 .maxlen         = sizeof(unsigned int),
 688                 .mode           = 0644,
 689                 .proc_handler   = &proc_dointvec_jiffies,
 690         },
 691         {
 692                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE,
 693                 .procname       = "ip_conntrack_tcp_timeout_close",
 694                 .data           = &ip_ct_tcp_timeout_close,
 695                 .maxlen         = sizeof(unsigned int),
 696                 .mode           = 0644,
 697                 .proc_handler   = &proc_dointvec_jiffies,
 698         },
 699         {
 700                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT,
 701                 .procname       = "ip_conntrack_udp_timeout",
 702                 .data           = &ip_ct_udp_timeout,
 703                 .maxlen         = sizeof(unsigned int),
 704                 .mode           = 0644,
 705                 .proc_handler   = &proc_dointvec_jiffies,
 706         },
 707         {
 708                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT_STREAM,
 709                 .procname       = "ip_conntrack_udp_timeout_stream",
 710                 .data           = &ip_ct_udp_timeout_stream,
 711                 .maxlen         = sizeof(unsigned int),
 712                 .mode           = 0644,
 713                 .proc_handler   = &proc_dointvec_jiffies,
 714         },
 715         {
 716                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 717                 .procname       = "ip_conntrack_icmp_timeout",
 718                 .data           = &ip_ct_icmp_timeout,
 719                 .maxlen         = sizeof(unsigned int),
 720                 .mode           = 0644,
 721                 .proc_handler   = &proc_dointvec_jiffies,
 722         },
 723         {
 724                 .ctl_name       = NET_IPV4_NF_CONNTRACK_GENERIC_TIMEOUT,
 725                 .procname       = "ip_conntrack_generic_timeout",
 726                 .data           = &ip_ct_generic_timeout,
 727                 .maxlen         = sizeof(unsigned int),
 728                 .mode           = 0644,
 729                 .proc_handler   = &proc_dointvec_jiffies,
 730         },
 731         {
 732                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 733                 .procname       = "ip_conntrack_log_invalid",
 734                 .data           = &ip_ct_log_invalid,
 735                 .maxlen         = sizeof(unsigned int),
 736                 .mode           = 0644,
 737                 .proc_handler   = &proc_dointvec_minmax,
 738                 .strategy       = &sysctl_intvec,
 739                 .extra1         = &log_invalid_proto_min,
 740                 .extra2         = &log_invalid_proto_max,
 741         },
 742         {
 743                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_MAX_RETRANS,
 744                 .procname       = "ip_conntrack_tcp_timeout_max_retrans",
 745                 .data           = &ip_ct_tcp_timeout_max_retrans,
 746                 .maxlen         = sizeof(unsigned int),
 747                 .mode           = 0644,
 748                 .proc_handler   = &proc_dointvec_jiffies,
 749         },
 750         {
 751                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_LOOSE,
 752                 .procname       = "ip_conntrack_tcp_loose",
 753                 .data           = &ip_ct_tcp_loose,
 754                 .maxlen         = sizeof(unsigned int),
 755                 .mode           = 0644,
 756                 .proc_handler   = &proc_dointvec,
 757         },
 758         {
 759                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_BE_LIBERAL,
 760                 .procname       = "ip_conntrack_tcp_be_liberal",
 761                 .data           = &ip_ct_tcp_be_liberal,
 762                 .maxlen         = sizeof(unsigned int),
 763                 .mode           = 0644,
 764                 .proc_handler   = &proc_dointvec,
 765         },
 766         {
 767                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_MAX_RETRANS,
 768                 .procname       = "ip_conntrack_tcp_max_retrans",
 769                 .data           = &ip_ct_tcp_max_retrans,
 770                 .maxlen         = sizeof(unsigned int),
 771                 .mode           = 0644,
 772                 .proc_handler   = &proc_dointvec,
 773         },
 774         { .ctl_name = 0 }
 775 };
 776
 777 #define NET_IP_CONNTRACK_MAX 2089
 778
 779 static ctl_table ip_ct_netfilter_table[] = {
 780         {
 781                 .ctl_name       = NET_IPV4_NETFILTER,
 782                 .procname       = "netfilter",
 783                 .mode           = 0555,
 784                 .child          = ip_ct_sysctl_table,
 785         },
 786         {
 787                 .ctl_name       = NET_IP_CONNTRACK_MAX,
 788                 .procname       = "ip_conntrack_max",
 789                 .data           = &ip_conntrack_max,
 790                 .maxlen         = sizeof(int),
 791                 .mode           = 0644,
 792                 .proc_handler   = &proc_dointvec
 793         },
 794         { .ctl_name = 0 }
 795 };
 796
 797 static ctl_table ip_ct_ipv4_table[] = {
 798         {
 799                 .ctl_name       = NET_IPV4,
 800                 .procname       = "ipv4",
 801                 .mode           = 0555,
 802                 .child          = ip_ct_netfilter_table,
 803         },
 804         { .ctl_name = 0 }
 805 };
 806
 807 static ctl_table ip_ct_net_table[] = {
 808         {
 809                 .ctl_name       = CTL_NET,
 810                 .procname       = "net",
 811                 .mode           = 0555,
 812                 .child          = ip_ct_ipv4_table,
 813         },
 814         { .ctl_name = 0 }
 815 };
 816
 817 EXPORT_SYMBOL(ip_ct_log_invalid);
 818 #endif /* CONFIG_SYSCTL */
 819
 820 static int init_or_cleanup(int init)
 821 {
 822 #ifdef CONFIG_PROC_FS
 823         struct proc_dir_entry *proc, *proc_exp, *proc_stat;
 824 #endif
 825         int ret = 0;
 826
 827         if (!init) goto cleanup;
 828
 829         ret = ip_conntrack_init();
 830         if (ret < 0)
 831                 goto cleanup_nothing;
 832
 833 #ifdef CONFIG_PROC_FS
 834         ret = -ENOMEM;
 835         proc = proc_net_fops_create("ip_conntrack", 0440, &ct_file_ops);
 836         if (!proc) goto cleanup_init;
 837
 838         proc_exp = proc_net_fops_create("ip_conntrack_expect", 0440,
 839                                         &exp_file_ops);
 840         if (!proc_exp) goto cleanup_proc;
 841
 842         proc_stat = create_proc_entry("ip_conntrack", S_IRUGO, proc_net_stat);
 843         if (!proc_stat)
 844                 goto cleanup_proc_exp;
 845
 846         proc_stat->proc_fops = &ct_cpu_seq_fops;
 847         proc_stat->owner = THIS_MODULE;
 848 #endif
 849
 850         ret = nf_register_hook(&ip_conntrack_defrag_ops);
 851         if (ret < 0) {
 852                 printk("ip_conntrack: can't register pre-routing defrag hook.\n");
 853                 goto cleanup_proc_stat;
 854         }
 855         ret = nf_register_hook(&ip_conntrack_defrag_local_out_ops);
 856         if (ret < 0) {
 857                 printk("ip_conntrack: can't register local_out defrag hook.\n");
 858                 goto cleanup_defragops;
 859         }
 860         ret = nf_register_hook(&ip_conntrack_in_ops);
 861         if (ret < 0) {
 862                 printk("ip_conntrack: can't register pre-routing hook.\n");
 863                 goto cleanup_defraglocalops;
 864         }
 865         ret = nf_register_hook(&ip_conntrack_local_out_ops);
 866         if (ret < 0) {
 867                 printk("ip_conntrack: can't register local out hook.\n");
 868                 goto cleanup_inops;
 869         }
 870         ret = nf_register_hook(&ip_conntrack_helper_in_ops);
 871         if (ret < 0) {
 872                 printk("ip_conntrack: can't register local in helper hook.\n");
 873                 goto cleanup_inandlocalops;
 874         }
 875         ret = nf_register_hook(&ip_conntrack_helper_out_ops);
 876         if (ret < 0) {
 877                 printk("ip_conntrack: can't register postrouting helper hook.\n");
 878                 goto cleanup_helperinops;
 879         }
 880         ret = nf_register_hook(&ip_conntrack_out_ops);
 881         if (ret < 0) {
 882                 printk("ip_conntrack: can't register post-routing hook.\n");
 883                 goto cleanup_helperoutops;
 884         }
 885         ret = nf_register_hook(&ip_conntrack_local_in_ops);
 886         if (ret < 0) {
 887                 printk("ip_conntrack: can't register local in hook.\n");
 888                 goto cleanup_inoutandlocalops;
 889         }
 890 #ifdef CONFIG_SYSCTL
 891         ip_ct_sysctl_header = register_sysctl_table(ip_ct_net_table, 0);
 892         if (ip_ct_sysctl_header == NULL) {
 893                 printk("ip_conntrack: can't register to sysctl.\n");
 894                 ret = -ENOMEM;
 895                 goto cleanup_localinops;
 896         }
 897 #endif
 898
 899         return ret;
 900
 901  cleanup:
 902 #ifdef CONFIG_SYSCTL
 903         unregister_sysctl_table(ip_ct_sysctl_header);
 904  cleanup_localinops:
 905 #endif
 906         nf_unregister_hook(&ip_conntrack_local_in_ops);
 907  cleanup_inoutandlocalops:
 908         nf_unregister_hook(&ip_conntrack_out_ops);
 909  cleanup_helperoutops:
 910         nf_unregister_hook(&ip_conntrack_helper_out_ops);
 911  cleanup_helperinops:
 912         nf_unregister_hook(&ip_conntrack_helper_in_ops);
 913  cleanup_inandlocalops:
 914         nf_unregister_hook(&ip_conntrack_local_out_ops);
 915  cleanup_inops:
 916         nf_unregister_hook(&ip_conntrack_in_ops);
 917  cleanup_defraglocalops:
 918         nf_unregister_hook(&ip_conntrack_defrag_local_out_ops);
 919  cleanup_defragops:
 920         nf_unregister_hook(&ip_conntrack_defrag_ops);
 921  cleanup_proc_stat:
 922 #ifdef CONFIG_PROC_FS
 923         remove_proc_entry("ip_conntrack", proc_net_stat);
 924  cleanup_proc_exp:
 925         proc_net_remove("ip_conntrack_expect");
 926  cleanup_proc:
 927         proc_net_remove("ip_conntrack");
 928  cleanup_init:
 929 #endif /* CONFIG_PROC_FS */
 930         ip_conntrack_cleanup();
 931  cleanup_nothing:
 932         return ret;
 933 }
 934
 935 /* FIXME: Allow NULL functions and sub in pointers to generic for
 936    them. --RR */
 937 int ip_conntrack_protocol_register(struct ip_conntrack_protocol *proto)
 938 {
 939         int ret = 0;
 940
 941         WRITE_LOCK(&ip_conntrack_lock);
 942         if (ip_ct_protos[proto->proto] != &ip_conntrack_generic_protocol) {
 943                 ret = -EBUSY;
 944                 goto out;
 945         }
 946         ip_ct_protos[proto->proto] = proto;
 947  out:
 948         WRITE_UNLOCK(&ip_conntrack_lock);
 949         return ret;
 950 }
 951
 952 void ip_conntrack_protocol_unregister(struct ip_conntrack_protocol *proto)
 953 {
 954         WRITE_LOCK(&ip_conntrack_lock);
 955         ip_ct_protos[proto->proto] = &ip_conntrack_generic_protocol;
 956         WRITE_UNLOCK(&ip_conntrack_lock);
 957
 958         /* Somebody could be still looking at the proto in bh. */
 959         synchronize_net();
 960
 961         /* Remove all contrack entries for this protocol */
 962         ip_ct_iterate_cleanup(kill_proto, &proto->proto);
 963 }
 964
 965 static int __init init(void)
 966 {
 967         return init_or_cleanup(1);
 968 }
 969
 970 static void __exit fini(void)
 971 {
 972         init_or_cleanup(0);
 973 }
 974
 975 module_init(init);
 976 module_exit(fini);
 977
 978 /* Some modules need us, but don't depend directly on any symbol.
 979    They should call this. */
 980 void need_ip_conntrack(void)
 981 {
 982 }
 983
 984 EXPORT_SYMBOL(ip_conntrack_protocol_register);
 985 EXPORT_SYMBOL(ip_conntrack_protocol_unregister);
 986 EXPORT_SYMBOL(ip_ct_get_tuple);
 987 EXPORT_SYMBOL(invert_tuplepr);
 988 EXPORT_SYMBOL(ip_conntrack_alter_reply);
 989 EXPORT_SYMBOL(ip_conntrack_destroyed);
 990 EXPORT_SYMBOL(need_ip_conntrack);
 991 EXPORT_SYMBOL(ip_conntrack_helper_register);
 992 EXPORT_SYMBOL(ip_conntrack_helper_unregister);
 993 EXPORT_SYMBOL(ip_ct_iterate_cleanup);
 994 EXPORT_SYMBOL(ip_ct_refresh_acct);
 995 EXPORT_SYMBOL(ip_ct_protos);
 996 EXPORT_SYMBOL(ip_ct_find_proto);
 997 EXPORT_SYMBOL(ip_conntrack_expect_alloc);
 998 EXPORT_SYMBOL(ip_conntrack_expect_free);
 999 EXPORT_SYMBOL(ip_conntrack_expect_related);
1000 EXPORT_SYMBOL(ip_conntrack_unexpect_related);
1001 EXPORT_SYMBOL(ip_conntrack_tuple_taken);
1002 EXPORT_SYMBOL(ip_ct_gather_frags);
1003 EXPORT_SYMBOL(ip_conntrack_htable_size);
1004 EXPORT_SYMBOL(ip_conntrack_lock);
1005 EXPORT_SYMBOL(ip_conntrack_hash);
1006 EXPORT_SYMBOL(ip_conntrack_untracked);
1007 EXPORT_SYMBOL_GPL(ip_conntrack_find_get);
1008 EXPORT_SYMBOL_GPL(__ip_conntrack_find);
1009 EXPORT_SYMBOL_GPL(__ip_conntrack_exp_find);
1010 EXPORT_SYMBOL_GPL(ip_conntrack_put);
1011 #ifdef CONFIG_IP_NF_NAT_NEEDED
1012 EXPORT_SYMBOL(ip_conntrack_tcp_update);
1013 #endif