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         /* Gather fragments. */
 446         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 447                 *pskb = ip_ct_gather_frags(*pskb,
 448                                            hooknum == NF_IP_PRE_ROUTING ?
 449                                            IP_DEFRAG_CONNTRACK_IN :
 450                                            IP_DEFRAG_CONNTRACK_OUT);
 451                 if (!*pskb)
 452                         return NF_STOLEN;
 453         }
 454         return NF_ACCEPT;
 455 }
 456
 457 static unsigned int ip_refrag(unsigned int hooknum,
 458                               struct sk_buff **pskb,
 459                               const struct net_device *in,
 460                               const struct net_device *out,
 461                               int (*okfn)(struct sk_buff *))
 462 {
 463         struct rtable *rt = (struct rtable *)(*pskb)->dst;
 464
 465         /* We've seen it coming out the other side: confirm */
 466         if (ip_confirm(hooknum, pskb, in, out, okfn) != NF_ACCEPT)
 467                 return NF_DROP;
 468
 469         /* Local packets are never produced too large for their
 470            interface.  We degfragment them at LOCAL_OUT, however,
 471            so we have to refragment them here. */
 472         if ((*pskb)->len > dst_mtu(&rt->u.dst) &&
 473             !skb_shinfo(*pskb)->tso_size) {
 474                 /* No hook can be after us, so this should be OK. */
 475                 ip_fragment(*pskb, okfn);
 476                 return NF_STOLEN;
 477         }
 478         return NF_ACCEPT;
 479 }
 480
 481 static unsigned int ip_conntrack_local(unsigned int hooknum,
 482                                        struct sk_buff **pskb,
 483                                        const struct net_device *in,
 484                                        const struct net_device *out,
 485                                        int (*okfn)(struct sk_buff *))
 486 {
 487         /* root is playing with raw sockets. */
 488         if ((*pskb)->len < sizeof(struct iphdr)
 489             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 490                 if (net_ratelimit())
 491                         printk("ipt_hook: happy cracking.\n");
 492                 return NF_ACCEPT;
 493         }
 494         return ip_conntrack_in(hooknum, pskb, in, out, okfn);
 495 }
 496
 497 /* Connection tracking may drop packets, but never alters them, so
 498    make it the first hook. */
 499 static struct nf_hook_ops ip_conntrack_defrag_ops = {
 500         .hook           = ip_conntrack_defrag,
 501         .owner          = THIS_MODULE,
 502         .pf             = PF_INET,
 503         .hooknum        = NF_IP_PRE_ROUTING,
 504         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 505 };
 506
 507 static struct nf_hook_ops ip_conntrack_in_ops = {
 508         .hook           = ip_conntrack_in,
 509         .owner          = THIS_MODULE,
 510         .pf             = PF_INET,
 511         .hooknum        = NF_IP_PRE_ROUTING,
 512         .priority       = NF_IP_PRI_CONNTRACK,
 513 };
 514
 515 static struct nf_hook_ops ip_conntrack_defrag_local_out_ops = {
 516         .hook           = ip_conntrack_defrag,
 517         .owner          = THIS_MODULE,
 518         .pf             = PF_INET,
 519         .hooknum        = NF_IP_LOCAL_OUT,
 520         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 521 };
 522
 523 static struct nf_hook_ops ip_conntrack_local_out_ops = {
 524         .hook           = ip_conntrack_local,
 525         .owner          = THIS_MODULE,
 526         .pf             = PF_INET,
 527         .hooknum        = NF_IP_LOCAL_OUT,
 528         .priority       = NF_IP_PRI_CONNTRACK,
 529 };
 530
 531 /* helpers */
 532 static struct nf_hook_ops ip_conntrack_helper_out_ops = {
 533         .hook           = ip_conntrack_help,
 534         .owner          = THIS_MODULE,
 535         .pf             = PF_INET,
 536         .hooknum        = NF_IP_POST_ROUTING,
 537         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 538 };
 539
 540 static struct nf_hook_ops ip_conntrack_helper_in_ops = {
 541         .hook           = ip_conntrack_help,
 542         .owner          = THIS_MODULE,
 543         .pf             = PF_INET,
 544         .hooknum        = NF_IP_LOCAL_IN,
 545         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 546 };
 547
 548 /* Refragmenter; last chance. */
 549 static struct nf_hook_ops ip_conntrack_out_ops = {
 550         .hook           = ip_refrag,
 551         .owner          = THIS_MODULE,
 552         .pf             = PF_INET,
 553         .hooknum        = NF_IP_POST_ROUTING,
 554         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 555 };
 556
 557 static struct nf_hook_ops ip_conntrack_local_in_ops = {
 558         .hook           = ip_confirm,
 559         .owner          = THIS_MODULE,
 560         .pf             = PF_INET,
 561         .hooknum        = NF_IP_LOCAL_IN,
 562         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 563 };
 564
 565 /* Sysctl support */
 566
 567 #ifdef CONFIG_SYSCTL
 568
 569 /* From ip_conntrack_core.c */
 570 extern int ip_conntrack_max;
 571 extern unsigned int ip_conntrack_htable_size;
 572
 573 /* From ip_conntrack_proto_tcp.c */
 574 extern unsigned long ip_ct_tcp_timeout_syn_sent;
 575 extern unsigned long ip_ct_tcp_timeout_syn_recv;
 576 extern unsigned long ip_ct_tcp_timeout_established;
 577 extern unsigned long ip_ct_tcp_timeout_fin_wait;
 578 extern unsigned long ip_ct_tcp_timeout_close_wait;
 579 extern unsigned long ip_ct_tcp_timeout_last_ack;
 580 extern unsigned long ip_ct_tcp_timeout_time_wait;
 581 extern unsigned long ip_ct_tcp_timeout_close;
 582 extern unsigned long ip_ct_tcp_timeout_max_retrans;
 583 extern int ip_ct_tcp_loose;
 584 extern int ip_ct_tcp_be_liberal;
 585 extern int ip_ct_tcp_max_retrans;
 586
 587 /* From ip_conntrack_proto_udp.c */
 588 extern unsigned long ip_ct_udp_timeout;
 589 extern unsigned long ip_ct_udp_timeout_stream;
 590
 591 /* From ip_conntrack_proto_icmp.c */
 592 extern unsigned long ip_ct_icmp_timeout;
 593
 594 /* From ip_conntrack_proto_icmp.c */
 595 extern unsigned long ip_ct_generic_timeout;
 596
 597 /* Log invalid packets of a given protocol */
 598 static int log_invalid_proto_min = 0;
 599 static int log_invalid_proto_max = 255;
 600
 601 static struct ctl_table_header *ip_ct_sysctl_header;
 602
 603 static ctl_table ip_ct_sysctl_table[] = {
 604         {
 605                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 606                 .procname       = "ip_conntrack_max",
 607                 .data           = &ip_conntrack_max,
 608                 .maxlen         = sizeof(int),
 609                 .mode           = 0644,
 610                 .proc_handler   = &proc_dointvec,
 611         },
 612         {
 613                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 614                 .procname       = "ip_conntrack_count",
 615                 .data           = &ip_conntrack_count,
 616                 .maxlen         = sizeof(int),
 617                 .mode           = 0444,
 618                 .proc_handler   = &proc_dointvec,
 619         },
 620         {
 621                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 622                 .procname       = "ip_conntrack_buckets",
 623                 .data           = &ip_conntrack_htable_size,
 624                 .maxlen         = sizeof(unsigned int),
 625                 .mode           = 0444,
 626                 .proc_handler   = &proc_dointvec,
 627         },
 628         {
 629                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_SENT,
 630                 .procname       = "ip_conntrack_tcp_timeout_syn_sent",
 631                 .data           = &ip_ct_tcp_timeout_syn_sent,
 632                 .maxlen         = sizeof(unsigned int),
 633                 .mode           = 0644,
 634                 .proc_handler   = &proc_dointvec_jiffies,
 635         },
 636         {
 637                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_RECV,
 638                 .procname       = "ip_conntrack_tcp_timeout_syn_recv",
 639                 .data           = &ip_ct_tcp_timeout_syn_recv,
 640                 .maxlen         = sizeof(unsigned int),
 641                 .mode           = 0644,
 642                 .proc_handler   = &proc_dointvec_jiffies,
 643         },
 644         {
 645                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_ESTABLISHED,
 646                 .procname       = "ip_conntrack_tcp_timeout_established",
 647                 .data           = &ip_ct_tcp_timeout_established,
 648                 .maxlen         = sizeof(unsigned int),
 649                 .mode           = 0644,
 650                 .proc_handler   = &proc_dointvec_jiffies,
 651         },
 652         {
 653                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_FIN_WAIT,
 654                 .procname       = "ip_conntrack_tcp_timeout_fin_wait",
 655                 .data           = &ip_ct_tcp_timeout_fin_wait,
 656                 .maxlen         = sizeof(unsigned int),
 657                 .mode           = 0644,
 658                 .proc_handler   = &proc_dointvec_jiffies,
 659         },
 660         {
 661                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE_WAIT,
 662                 .procname       = "ip_conntrack_tcp_timeout_close_wait",
 663                 .data           = &ip_ct_tcp_timeout_close_wait,
 664                 .maxlen         = sizeof(unsigned int),
 665                 .mode           = 0644,
 666                 .proc_handler   = &proc_dointvec_jiffies,
 667         },
 668         {
 669                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_LAST_ACK,
 670                 .procname       = "ip_conntrack_tcp_timeout_last_ack",
 671                 .data           = &ip_ct_tcp_timeout_last_ack,
 672                 .maxlen         = sizeof(unsigned int),
 673                 .mode           = 0644,
 674                 .proc_handler   = &proc_dointvec_jiffies,
 675         },
 676         {
 677                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_TIME_WAIT,
 678                 .procname       = "ip_conntrack_tcp_timeout_time_wait",
 679                 .data           = &ip_ct_tcp_timeout_time_wait,
 680                 .maxlen         = sizeof(unsigned int),
 681                 .mode           = 0644,
 682                 .proc_handler   = &proc_dointvec_jiffies,
 683         },
 684         {
 685                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE,
 686                 .procname       = "ip_conntrack_tcp_timeout_close",
 687                 .data           = &ip_ct_tcp_timeout_close,
 688                 .maxlen         = sizeof(unsigned int),
 689                 .mode           = 0644,
 690                 .proc_handler   = &proc_dointvec_jiffies,
 691         },
 692         {
 693                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT,
 694                 .procname       = "ip_conntrack_udp_timeout",
 695                 .data           = &ip_ct_udp_timeout,
 696                 .maxlen         = sizeof(unsigned int),
 697                 .mode           = 0644,
 698                 .proc_handler   = &proc_dointvec_jiffies,
 699         },
 700         {
 701                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT_STREAM,
 702                 .procname       = "ip_conntrack_udp_timeout_stream",
 703                 .data           = &ip_ct_udp_timeout_stream,
 704                 .maxlen         = sizeof(unsigned int),
 705                 .mode           = 0644,
 706                 .proc_handler   = &proc_dointvec_jiffies,
 707         },
 708         {
 709                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 710                 .procname       = "ip_conntrack_icmp_timeout",
 711                 .data           = &ip_ct_icmp_timeout,
 712                 .maxlen         = sizeof(unsigned int),
 713                 .mode           = 0644,
 714                 .proc_handler   = &proc_dointvec_jiffies,
 715         },
 716         {
 717                 .ctl_name       = NET_IPV4_NF_CONNTRACK_GENERIC_TIMEOUT,
 718                 .procname       = "ip_conntrack_generic_timeout",
 719                 .data           = &ip_ct_generic_timeout,
 720                 .maxlen         = sizeof(unsigned int),
 721                 .mode           = 0644,
 722                 .proc_handler   = &proc_dointvec_jiffies,
 723         },
 724         {
 725                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 726                 .procname       = "ip_conntrack_log_invalid",
 727                 .data           = &ip_ct_log_invalid,
 728                 .maxlen         = sizeof(unsigned int),
 729                 .mode           = 0644,
 730                 .proc_handler   = &proc_dointvec_minmax,
 731                 .strategy       = &sysctl_intvec,
 732                 .extra1         = &log_invalid_proto_min,
 733                 .extra2         = &log_invalid_proto_max,
 734         },
 735         {
 736                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_MAX_RETRANS,
 737                 .procname       = "ip_conntrack_tcp_timeout_max_retrans",
 738                 .data           = &ip_ct_tcp_timeout_max_retrans,
 739                 .maxlen         = sizeof(unsigned int),
 740                 .mode           = 0644,
 741                 .proc_handler   = &proc_dointvec_jiffies,
 742         },
 743         {
 744                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_LOOSE,
 745                 .procname       = "ip_conntrack_tcp_loose",
 746                 .data           = &ip_ct_tcp_loose,
 747                 .maxlen         = sizeof(unsigned int),
 748                 .mode           = 0644,
 749                 .proc_handler   = &proc_dointvec,
 750         },
 751         {
 752                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_BE_LIBERAL,
 753                 .procname       = "ip_conntrack_tcp_be_liberal",
 754                 .data           = &ip_ct_tcp_be_liberal,
 755                 .maxlen         = sizeof(unsigned int),
 756                 .mode           = 0644,
 757                 .proc_handler   = &proc_dointvec,
 758         },
 759         {
 760                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_MAX_RETRANS,
 761                 .procname       = "ip_conntrack_tcp_max_retrans",
 762                 .data           = &ip_ct_tcp_max_retrans,
 763                 .maxlen         = sizeof(unsigned int),
 764                 .mode           = 0644,
 765                 .proc_handler   = &proc_dointvec,
 766         },
 767         { .ctl_name = 0 }
 768 };
 769
 770 #define NET_IP_CONNTRACK_MAX 2089
 771
 772 static ctl_table ip_ct_netfilter_table[] = {
 773         {
 774                 .ctl_name       = NET_IPV4_NETFILTER,
 775                 .procname       = "netfilter",
 776                 .mode           = 0555,
 777                 .child          = ip_ct_sysctl_table,
 778         },
 779         {
 780                 .ctl_name       = NET_IP_CONNTRACK_MAX,
 781                 .procname       = "ip_conntrack_max",
 782                 .data           = &ip_conntrack_max,
 783                 .maxlen         = sizeof(int),
 784                 .mode           = 0644,
 785                 .proc_handler   = &proc_dointvec
 786         },
 787         { .ctl_name = 0 }
 788 };
 789
 790 static ctl_table ip_ct_ipv4_table[] = {
 791         {
 792                 .ctl_name       = NET_IPV4,
 793                 .procname       = "ipv4",
 794                 .mode           = 0555,
 795                 .child          = ip_ct_netfilter_table,
 796         },
 797         { .ctl_name = 0 }
 798 };
 799
 800 static ctl_table ip_ct_net_table[] = {
 801         {
 802                 .ctl_name       = CTL_NET,
 803                 .procname       = "net",
 804                 .mode           = 0555,
 805                 .child          = ip_ct_ipv4_table,
 806         },
 807         { .ctl_name = 0 }
 808 };
 809
 810 EXPORT_SYMBOL(ip_ct_log_invalid);
 811 #endif /* CONFIG_SYSCTL */
 812
 813 static int init_or_cleanup(int init)
 814 {
 815 #ifdef CONFIG_PROC_FS
 816         struct proc_dir_entry *proc, *proc_exp, *proc_stat;
 817 #endif
 818         int ret = 0;
 819
 820         if (!init) goto cleanup;
 821
 822         ret = ip_conntrack_init();
 823         if (ret < 0)
 824                 goto cleanup_nothing;
 825
 826 #ifdef CONFIG_PROC_FS
 827         ret = -ENOMEM;
 828         proc = proc_net_fops_create("ip_conntrack", 0440, &ct_file_ops);
 829         if (!proc) goto cleanup_init;
 830
 831         proc_exp = proc_net_fops_create("ip_conntrack_expect", 0440,
 832                                         &exp_file_ops);
 833         if (!proc_exp) goto cleanup_proc;
 834
 835         proc_stat = create_proc_entry("ip_conntrack", S_IRUGO, proc_net_stat);
 836         if (!proc_stat)
 837                 goto cleanup_proc_exp;
 838
 839         proc_stat->proc_fops = &ct_cpu_seq_fops;
 840         proc_stat->owner = THIS_MODULE;
 841 #endif
 842
 843         ret = nf_register_hook(&ip_conntrack_defrag_ops);
 844         if (ret < 0) {
 845                 printk("ip_conntrack: can't register pre-routing defrag hook.\n");
 846                 goto cleanup_proc_stat;
 847         }
 848         ret = nf_register_hook(&ip_conntrack_defrag_local_out_ops);
 849         if (ret < 0) {
 850                 printk("ip_conntrack: can't register local_out defrag hook.\n");
 851                 goto cleanup_defragops;
 852         }
 853         ret = nf_register_hook(&ip_conntrack_in_ops);
 854         if (ret < 0) {
 855                 printk("ip_conntrack: can't register pre-routing hook.\n");
 856                 goto cleanup_defraglocalops;
 857         }
 858         ret = nf_register_hook(&ip_conntrack_local_out_ops);
 859         if (ret < 0) {
 860                 printk("ip_conntrack: can't register local out hook.\n");
 861                 goto cleanup_inops;
 862         }
 863         ret = nf_register_hook(&ip_conntrack_helper_in_ops);
 864         if (ret < 0) {
 865                 printk("ip_conntrack: can't register local in helper hook.\n");
 866                 goto cleanup_inandlocalops;
 867         }
 868         ret = nf_register_hook(&ip_conntrack_helper_out_ops);
 869         if (ret < 0) {
 870                 printk("ip_conntrack: can't register postrouting helper hook.\n");
 871                 goto cleanup_helperinops;
 872         }
 873         ret = nf_register_hook(&ip_conntrack_out_ops);
 874         if (ret < 0) {
 875                 printk("ip_conntrack: can't register post-routing hook.\n");
 876                 goto cleanup_helperoutops;
 877         }
 878         ret = nf_register_hook(&ip_conntrack_local_in_ops);
 879         if (ret < 0) {
 880                 printk("ip_conntrack: can't register local in hook.\n");
 881                 goto cleanup_inoutandlocalops;
 882         }
 883 #ifdef CONFIG_SYSCTL
 884         ip_ct_sysctl_header = register_sysctl_table(ip_ct_net_table, 0);
 885         if (ip_ct_sysctl_header == NULL) {
 886                 printk("ip_conntrack: can't register to sysctl.\n");
 887                 ret = -ENOMEM;
 888                 goto cleanup_localinops;
 889         }
 890 #endif
 891
 892         return ret;
 893
 894  cleanup:
 895 #ifdef CONFIG_SYSCTL
 896         unregister_sysctl_table(ip_ct_sysctl_header);
 897  cleanup_localinops:
 898 #endif
 899         nf_unregister_hook(&ip_conntrack_local_in_ops);
 900  cleanup_inoutandlocalops:
 901         nf_unregister_hook(&ip_conntrack_out_ops);
 902  cleanup_helperoutops:
 903         nf_unregister_hook(&ip_conntrack_helper_out_ops);
 904  cleanup_helperinops:
 905         nf_unregister_hook(&ip_conntrack_helper_in_ops);
 906  cleanup_inandlocalops:
 907         nf_unregister_hook(&ip_conntrack_local_out_ops);
 908  cleanup_inops:
 909         nf_unregister_hook(&ip_conntrack_in_ops);
 910  cleanup_defraglocalops:
 911         nf_unregister_hook(&ip_conntrack_defrag_local_out_ops);
 912  cleanup_defragops:
 913         nf_unregister_hook(&ip_conntrack_defrag_ops);
 914  cleanup_proc_stat:
 915 #ifdef CONFIG_PROC_FS
 916         remove_proc_entry("ip_conntrack", proc_net_stat);
 917  cleanup_proc_exp:
 918         proc_net_remove("ip_conntrack_expect");
 919  cleanup_proc:
 920         proc_net_remove("ip_conntrack");
 921  cleanup_init:
 922 #endif /* CONFIG_PROC_FS */
 923         ip_conntrack_cleanup();
 924  cleanup_nothing:
 925         return ret;
 926 }
 927
 928 /* FIXME: Allow NULL functions and sub in pointers to generic for
 929    them. --RR */
 930 int ip_conntrack_protocol_register(struct ip_conntrack_protocol *proto)
 931 {
 932         int ret = 0;
 933
 934         WRITE_LOCK(&ip_conntrack_lock);
 935         if (ip_ct_protos[proto->proto] != &ip_conntrack_generic_protocol) {
 936                 ret = -EBUSY;
 937                 goto out;
 938         }
 939         ip_ct_protos[proto->proto] = proto;
 940  out:
 941         WRITE_UNLOCK(&ip_conntrack_lock);
 942         return ret;
 943 }
 944
 945 void ip_conntrack_protocol_unregister(struct ip_conntrack_protocol *proto)
 946 {
 947         WRITE_LOCK(&ip_conntrack_lock);
 948         ip_ct_protos[proto->proto] = &ip_conntrack_generic_protocol;
 949         WRITE_UNLOCK(&ip_conntrack_lock);
 950
 951         /* Somebody could be still looking at the proto in bh. */
 952         synchronize_net();
 953
 954         /* Remove all contrack entries for this protocol */
 955         ip_ct_iterate_cleanup(kill_proto, &proto->proto);
 956 }
 957
 958 static int __init init(void)
 959 {
 960         return init_or_cleanup(1);
 961 }
 962
 963 static void __exit fini(void)
 964 {
 965         init_or_cleanup(0);
 966 }
 967
 968 module_init(init);
 969 module_exit(fini);
 970
 971 /* Some modules need us, but don't depend directly on any symbol.
 972    They should call this. */
 973 void need_ip_conntrack(void)
 974 {
 975 }
 976
 977 EXPORT_SYMBOL(ip_conntrack_protocol_register);
 978 EXPORT_SYMBOL(ip_conntrack_protocol_unregister);
 979 EXPORT_SYMBOL(ip_ct_get_tuple);
 980 EXPORT_SYMBOL(invert_tuplepr);
 981 EXPORT_SYMBOL(ip_conntrack_alter_reply);
 982 EXPORT_SYMBOL(ip_conntrack_destroyed);
 983 EXPORT_SYMBOL(need_ip_conntrack);
 984 EXPORT_SYMBOL(ip_conntrack_helper_register);
 985 EXPORT_SYMBOL(ip_conntrack_helper_unregister);
 986 EXPORT_SYMBOL(ip_ct_iterate_cleanup);
 987 EXPORT_SYMBOL(ip_ct_refresh_acct);
 988 EXPORT_SYMBOL(ip_ct_protos);
 989 EXPORT_SYMBOL(ip_ct_find_proto);
 990 EXPORT_SYMBOL(ip_conntrack_expect_alloc);
 991 EXPORT_SYMBOL(ip_conntrack_expect_free);
 992 EXPORT_SYMBOL(ip_conntrack_expect_related);
 993 EXPORT_SYMBOL(ip_conntrack_unexpect_related);
 994 EXPORT_SYMBOL(ip_conntrack_tuple_taken);
 995 EXPORT_SYMBOL(ip_ct_gather_frags);
 996 EXPORT_SYMBOL(ip_conntrack_htable_size);
 997 EXPORT_SYMBOL(ip_conntrack_lock);
 998 EXPORT_SYMBOL(ip_conntrack_hash);
 999 EXPORT_SYMBOL(ip_conntrack_untracked);
1000 EXPORT_SYMBOL_GPL(ip_conntrack_find_get);
1001 EXPORT_SYMBOL_GPL(ip_conntrack_put);
1002 #ifdef CONFIG_IP_NF_NAT_NEEDED
1003 EXPORT_SYMBOL(ip_conntrack_tcp_update);
1004 #endif