net/ipv4/netfilter/ip_conntrack_standalone.c

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