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 (seq_printf(s, "use=%u\n", atomic_read(&conntrack->ct_general.use)))
 160                 return 1;
 161
 162         return 0;
 163 }
 164
 165 static int ct_seq_show(struct seq_file *s, void *v)
 166 {
 167         struct list_head *list = v;
 168         int ret = 0;
 169
 170         /* FIXME: Simply truncates if hash chain too long. */
 171         READ_LOCK(&ip_conntrack_lock);
 172         if (LIST_FIND(list, ct_seq_real_show,
 173                       struct ip_conntrack_tuple_hash *, s))
 174                 ret = -ENOSPC;
 175         READ_UNLOCK(&ip_conntrack_lock);
 176         return ret;
 177 }
 178
 179 static struct seq_operations ct_seq_ops = {
 180         .start = ct_seq_start,
 181         .next  = ct_seq_next,
 182         .stop  = ct_seq_stop,
 183         .show  = ct_seq_show
 184 };
 185
 186 static int ct_open(struct inode *inode, struct file *file)
 187 {
 188         return seq_open(file, &ct_seq_ops);
 189 }
 190
 191 static struct file_operations ct_file_ops = {
 192         .owner   = THIS_MODULE,
 193         .open    = ct_open,
 194         .read    = seq_read,
 195         .llseek  = seq_lseek,
 196         .release = seq_release
 197 };
 198
 199 /* expects */
 200 static void *exp_seq_start(struct seq_file *s, loff_t *pos)
 201 {
 202         struct list_head *e = &ip_conntrack_expect_list;
 203         loff_t i;
 204
 205         /* strange seq_file api calls stop even if we fail,
 206          * thus we need to grab lock since stop unlocks */
 207         READ_LOCK(&ip_conntrack_lock);
 208         READ_LOCK(&ip_conntrack_expect_tuple_lock);
 209
 210         if (list_empty(e))
 211                 return NULL;
 212
 213         for (i = 0; i <= *pos; i++) {
 214                 e = e->next;
 215                 if (e == &ip_conntrack_expect_list)
 216                         return NULL;
 217         }
 218         return e;
 219 }
 220
 221 static void *exp_seq_next(struct seq_file *s, void *v, loff_t *pos)
 222 {
 223         struct list_head *e = v;
 224
 225         e = e->next;
 226
 227         if (e == &ip_conntrack_expect_list)
 228                 return NULL;
 229
 230         return e;
 231 }
 232
 233 static void exp_seq_stop(struct seq_file *s, void *v)
 234 {
 235         READ_UNLOCK(&ip_conntrack_expect_tuple_lock);
 236         READ_UNLOCK(&ip_conntrack_lock);
 237 }
 238
 239 static int exp_seq_show(struct seq_file *s, void *v)
 240 {
 241         struct ip_conntrack_expect *expect = v;
 242
 243         if (expect->expectant->helper->timeout)
 244                 seq_printf(s, "%lu ", timer_pending(&expect->timeout)
 245                            ? (expect->timeout.expires - jiffies)/HZ : 0);
 246         else
 247                 seq_printf(s, "- ");
 248
 249         seq_printf(s, "use=%u proto=%u ", atomic_read(&expect->use),
 250                    expect->tuple.dst.protonum);
 251
 252         print_tuple(s, &expect->tuple,
 253                     ip_ct_find_proto(expect->tuple.dst.protonum));
 254         return seq_putc(s, '\n');
 255 }
 256
 257 static struct seq_operations exp_seq_ops = {
 258         .start = exp_seq_start,
 259         .next = exp_seq_next,
 260         .stop = exp_seq_stop,
 261         .show = exp_seq_show
 262 };
 263
 264 static int exp_open(struct inode *inode, struct file *file)
 265 {
 266         return seq_open(file, &exp_seq_ops);
 267 }
 268
 269 static struct file_operations exp_file_ops = {
 270         .owner   = THIS_MODULE,
 271         .open    = exp_open,
 272         .read    = seq_read,
 273         .llseek  = seq_lseek,
 274         .release = seq_release
 275 };
 276
 277 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 278 {
 279         int cpu;
 280
 281         if (*pos == 0)
 282                 return SEQ_START_TOKEN;
 283
 284         for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
 285                 if (!cpu_possible(cpu))
 286                         continue;
 287                 *pos = cpu+1;
 288                 return &per_cpu(ip_conntrack_stat, cpu);
 289         }
 290
 291         return NULL;
 292 }
 293
 294 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 295 {
 296         int cpu;
 297
 298         for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
 299                 if (!cpu_possible(cpu))
 300                         continue;
 301                 *pos = cpu+1;
 302                 return &per_cpu(ip_conntrack_stat, cpu);
 303         }
 304
 305         return NULL;
 306 }
 307
 308 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 309 {
 310 }
 311
 312 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 313 {
 314         unsigned int nr_conntracks = atomic_read(&ip_conntrack_count);
 315         struct ip_conntrack_stat *st = v;
 316
 317         if (v == SEQ_START_TOKEN) {
 318                 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");
 319                 return 0;
 320         }
 321
 322         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 323                         "%08x %08x %08x %08x %08x  %08x %08x %08x \n",
 324                    nr_conntracks,
 325                    st->searched,
 326                    st->found,
 327                    st->new,
 328                    st->invalid,
 329                    st->ignore,
 330                    st->delete,
 331                    st->delete_list,
 332                    st->insert,
 333                    st->insert_failed,
 334                    st->drop,
 335                    st->early_drop,
 336                    st->error,
 337
 338                    st->expect_new,
 339                    st->expect_create,
 340                    st->expect_delete
 341                 );
 342         return 0;
 343 }
 344
 345 static struct seq_operations ct_cpu_seq_ops = {
 346         .start  = ct_cpu_seq_start,
 347         .next   = ct_cpu_seq_next,
 348         .stop   = ct_cpu_seq_stop,
 349         .show   = ct_cpu_seq_show,
 350 };
 351
 352 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 353 {
 354         return seq_open(file, &ct_cpu_seq_ops);
 355 }
 356
 357 static struct file_operations ct_cpu_seq_fops = {
 358         .owner   = THIS_MODULE,
 359         .open    = ct_cpu_seq_open,
 360         .read    = seq_read,
 361         .llseek  = seq_lseek,
 362         .release = seq_release_private,
 363 };
 364 #endif
 365
 366 static unsigned int ip_confirm(unsigned int hooknum,
 367                                struct sk_buff **pskb,
 368                                const struct net_device *in,
 369                                const struct net_device *out,
 370                                int (*okfn)(struct sk_buff *))
 371 {
 372         /* We've seen it coming out the other side: confirm it */
 373         return ip_conntrack_confirm(*pskb);
 374 }
 375
 376 static unsigned int ip_conntrack_defrag(unsigned int hooknum,
 377                                         struct sk_buff **pskb,
 378                                         const struct net_device *in,
 379                                         const struct net_device *out,
 380                                         int (*okfn)(struct sk_buff *))
 381 {
 382         /* Previously seen (loopback)?  Ignore.  Do this before
 383            fragment check. */
 384         if ((*pskb)->nfct)
 385                 return NF_ACCEPT;
 386
 387         /* Gather fragments. */
 388         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 389                 *pskb = ip_ct_gather_frags(*pskb);
 390                 if (!*pskb)
 391                         return NF_STOLEN;
 392         }
 393         return NF_ACCEPT;
 394 }
 395
 396 static unsigned int ip_refrag(unsigned int hooknum,
 397                               struct sk_buff **pskb,
 398                               const struct net_device *in,
 399                               const struct net_device *out,
 400                               int (*okfn)(struct sk_buff *))
 401 {
 402         struct rtable *rt = (struct rtable *)(*pskb)->dst;
 403
 404         /* We've seen it coming out the other side: confirm */
 405         if (ip_confirm(hooknum, pskb, in, out, okfn) != NF_ACCEPT)
 406                 return NF_DROP;
 407
 408         /* Local packets are never produced too large for their
 409            interface.  We degfragment them at LOCAL_OUT, however,
 410            so we have to refragment them here. */
 411         if ((*pskb)->len > dst_pmtu(&rt->u.dst) &&
 412             !skb_shinfo(*pskb)->tso_size) {
 413                 /* No hook can be after us, so this should be OK. */
 414                 ip_fragment(*pskb, okfn);
 415                 return NF_STOLEN;
 416         }
 417         return NF_ACCEPT;
 418 }
 419
 420 static unsigned int ip_conntrack_local(unsigned int hooknum,
 421                                        struct sk_buff **pskb,
 422                                        const struct net_device *in,
 423                                        const struct net_device *out,
 424                                        int (*okfn)(struct sk_buff *))
 425 {
 426         /* root is playing with raw sockets. */
 427         if ((*pskb)->len < sizeof(struct iphdr)
 428             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 429                 if (net_ratelimit())
 430                         printk("ipt_hook: happy cracking.\n");
 431                 return NF_ACCEPT;
 432         }
 433         return ip_conntrack_in(hooknum, pskb, in, out, okfn);
 434 }
 435
 436 /* Connection tracking may drop packets, but never alters them, so
 437    make it the first hook. */
 438 static struct nf_hook_ops ip_conntrack_defrag_ops = {
 439         .hook           = ip_conntrack_defrag,
 440         .owner          = THIS_MODULE,
 441         .pf             = PF_INET,
 442         .hooknum        = NF_IP_PRE_ROUTING,
 443         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 444 };
 445
 446 static struct nf_hook_ops ip_conntrack_in_ops = {
 447         .hook           = ip_conntrack_in,
 448         .owner          = THIS_MODULE,
 449         .pf             = PF_INET,
 450         .hooknum        = NF_IP_PRE_ROUTING,
 451         .priority       = NF_IP_PRI_CONNTRACK,
 452 };
 453
 454 static struct nf_hook_ops ip_conntrack_defrag_local_out_ops = {
 455         .hook           = ip_conntrack_defrag,
 456         .owner          = THIS_MODULE,
 457         .pf             = PF_INET,
 458         .hooknum        = NF_IP_LOCAL_OUT,
 459         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 460 };
 461
 462 static struct nf_hook_ops ip_conntrack_local_out_ops = {
 463         .hook           = ip_conntrack_local,
 464         .owner          = THIS_MODULE,
 465         .pf             = PF_INET,
 466         .hooknum        = NF_IP_LOCAL_OUT,
 467         .priority       = NF_IP_PRI_CONNTRACK,
 468 };
 469
 470 /* Refragmenter; last chance. */
 471 static struct nf_hook_ops ip_conntrack_out_ops = {
 472         .hook           = ip_refrag,
 473         .owner          = THIS_MODULE,
 474         .pf             = PF_INET,
 475         .hooknum        = NF_IP_POST_ROUTING,
 476         .priority       = NF_IP_PRI_LAST,
 477 };
 478
 479 static struct nf_hook_ops ip_conntrack_local_in_ops = {
 480         .hook           = ip_confirm,
 481         .owner          = THIS_MODULE,
 482         .pf             = PF_INET,
 483         .hooknum        = NF_IP_LOCAL_IN,
 484         .priority       = NF_IP_PRI_LAST-1,
 485 };
 486
 487 /* Sysctl support */
 488
 489 #ifdef CONFIG_SYSCTL
 490
 491 /* From ip_conntrack_core.c */
 492 extern int ip_conntrack_max;
 493 extern unsigned int ip_conntrack_htable_size;
 494
 495 /* From ip_conntrack_proto_tcp.c */
 496 extern unsigned long ip_ct_tcp_timeout_syn_sent;
 497 extern unsigned long ip_ct_tcp_timeout_syn_recv;
 498 extern unsigned long ip_ct_tcp_timeout_established;
 499 extern unsigned long ip_ct_tcp_timeout_fin_wait;
 500 extern unsigned long ip_ct_tcp_timeout_close_wait;
 501 extern unsigned long ip_ct_tcp_timeout_last_ack;
 502 extern unsigned long ip_ct_tcp_timeout_time_wait;
 503 extern unsigned long ip_ct_tcp_timeout_close;
 504 extern unsigned long ip_ct_tcp_timeout_max_retrans;
 505 extern int ip_ct_tcp_loose;
 506 extern int ip_ct_tcp_be_liberal;
 507 extern int ip_ct_tcp_max_retrans;
 508
 509 /* From ip_conntrack_proto_udp.c */
 510 extern unsigned long ip_ct_udp_timeout;
 511 extern unsigned long ip_ct_udp_timeout_stream;
 512
 513 /* From ip_conntrack_proto_icmp.c */
 514 extern unsigned long ip_ct_icmp_timeout;
 515
 516 /* From ip_conntrack_proto_icmp.c */
 517 extern unsigned long ip_ct_generic_timeout;
 518
 519 /* Log invalid packets of a given protocol */
 520 static int log_invalid_proto_min = 0;
 521 static int log_invalid_proto_max = 255;
 522
 523 static struct ctl_table_header *ip_ct_sysctl_header;
 524
 525 static ctl_table ip_ct_sysctl_table[] = {
 526         {
 527                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 528                 .procname       = "ip_conntrack_max",
 529                 .data           = &ip_conntrack_max,
 530                 .maxlen         = sizeof(int),
 531                 .mode           = 0644,
 532                 .proc_handler   = &proc_dointvec,
 533         },
 534         {
 535                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 536                 .procname       = "ip_conntrack_count",
 537                 .data           = &ip_conntrack_count,
 538                 .maxlen         = sizeof(int),
 539                 .mode           = 0444,
 540                 .proc_handler   = &proc_dointvec,
 541         },
 542         {
 543                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 544                 .procname       = "ip_conntrack_buckets",
 545                 .data           = &ip_conntrack_htable_size,
 546                 .maxlen         = sizeof(unsigned int),
 547                 .mode           = 0444,
 548                 .proc_handler   = &proc_dointvec,
 549         },
 550         {
 551                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_SENT,
 552                 .procname       = "ip_conntrack_tcp_timeout_syn_sent",
 553                 .data           = &ip_ct_tcp_timeout_syn_sent,
 554                 .maxlen         = sizeof(unsigned int),
 555                 .mode           = 0644,
 556                 .proc_handler   = &proc_dointvec_jiffies,
 557         },
 558         {
 559                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_RECV,
 560                 .procname       = "ip_conntrack_tcp_timeout_syn_recv",
 561                 .data           = &ip_ct_tcp_timeout_syn_recv,
 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_ESTABLISHED,
 568                 .procname       = "ip_conntrack_tcp_timeout_established",
 569                 .data           = &ip_ct_tcp_timeout_established,
 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_FIN_WAIT,
 576                 .procname       = "ip_conntrack_tcp_timeout_fin_wait",
 577                 .data           = &ip_ct_tcp_timeout_fin_wait,
 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_CLOSE_WAIT,
 584                 .procname       = "ip_conntrack_tcp_timeout_close_wait",
 585                 .data           = &ip_ct_tcp_timeout_close_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_LAST_ACK,
 592                 .procname       = "ip_conntrack_tcp_timeout_last_ack",
 593                 .data           = &ip_ct_tcp_timeout_last_ack,
 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_TIME_WAIT,
 600                 .procname       = "ip_conntrack_tcp_timeout_time_wait",
 601                 .data           = &ip_ct_tcp_timeout_time_wait,
 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_CLOSE,
 608                 .procname       = "ip_conntrack_tcp_timeout_close",
 609                 .data           = &ip_ct_tcp_timeout_close,
 610                 .maxlen         = sizeof(unsigned int),
 611                 .mode           = 0644,
 612                 .proc_handler   = &proc_dointvec_jiffies,
 613         },
 614         {
 615                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT,
 616                 .procname       = "ip_conntrack_udp_timeout",
 617                 .data           = &ip_ct_udp_timeout,
 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_STREAM,
 624                 .procname       = "ip_conntrack_udp_timeout_stream",
 625                 .data           = &ip_ct_udp_timeout_stream,
 626                 .maxlen         = sizeof(unsigned int),
 627                 .mode           = 0644,
 628                 .proc_handler   = &proc_dointvec_jiffies,
 629         },
 630         {
 631                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 632                 .procname       = "ip_conntrack_icmp_timeout",
 633                 .data           = &ip_ct_icmp_timeout,
 634                 .maxlen         = sizeof(unsigned int),
 635                 .mode           = 0644,
 636                 .proc_handler   = &proc_dointvec_jiffies,
 637         },
 638         {
 639                 .ctl_name       = NET_IPV4_NF_CONNTRACK_GENERIC_TIMEOUT,
 640                 .procname       = "ip_conntrack_generic_timeout",
 641                 .data           = &ip_ct_generic_timeout,
 642                 .maxlen         = sizeof(unsigned int),
 643                 .mode           = 0644,
 644                 .proc_handler   = &proc_dointvec_jiffies,
 645         },
 646         {
 647                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 648                 .procname       = "ip_conntrack_log_invalid",
 649                 .data           = &ip_ct_log_invalid,
 650                 .maxlen         = sizeof(unsigned int),
 651                 .mode           = 0644,
 652                 .proc_handler   = &proc_dointvec_minmax,
 653                 .strategy       = &sysctl_intvec,
 654                 .extra1         = &log_invalid_proto_min,
 655                 .extra2         = &log_invalid_proto_max,
 656         },
 657         {
 658                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_MAX_RETRANS,
 659                 .procname       = "ip_conntrack_tcp_timeout_max_retrans",
 660                 .data           = &ip_ct_tcp_timeout_max_retrans,
 661                 .maxlen         = sizeof(unsigned int),
 662                 .mode           = 0644,
 663                 .proc_handler   = &proc_dointvec_jiffies,
 664         },
 665         {
 666                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_LOOSE,
 667                 .procname       = "ip_conntrack_tcp_loose",
 668                 .data           = &ip_ct_tcp_loose,
 669                 .maxlen         = sizeof(unsigned int),
 670                 .mode           = 0644,
 671                 .proc_handler   = &proc_dointvec,
 672         },
 673         {
 674                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_BE_LIBERAL,
 675                 .procname       = "ip_conntrack_tcp_be_liberal",
 676                 .data           = &ip_ct_tcp_be_liberal,
 677                 .maxlen         = sizeof(unsigned int),
 678                 .mode           = 0644,
 679                 .proc_handler   = &proc_dointvec,
 680         },
 681         {
 682                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_MAX_RETRANS,
 683                 .procname       = "ip_conntrack_tcp_max_retrans",
 684                 .data           = &ip_ct_tcp_max_retrans,
 685                 .maxlen         = sizeof(unsigned int),
 686                 .mode           = 0644,
 687                 .proc_handler   = &proc_dointvec,
 688         },
 689         { .ctl_name = 0 }
 690 };
 691
 692 #define NET_IP_CONNTRACK_MAX 2089
 693
 694 static ctl_table ip_ct_netfilter_table[] = {
 695         {
 696                 .ctl_name       = NET_IPV4_NETFILTER,
 697                 .procname       = "netfilter",
 698                 .mode           = 0555,
 699                 .child          = ip_ct_sysctl_table,
 700         },
 701         {
 702                 .ctl_name       = NET_IP_CONNTRACK_MAX,
 703                 .procname       = "ip_conntrack_max",
 704                 .data           = &ip_conntrack_max,
 705                 .maxlen         = sizeof(int),
 706                 .mode           = 0644,
 707                 .proc_handler   = &proc_dointvec
 708         },
 709         { .ctl_name = 0 }
 710 };
 711
 712 static ctl_table ip_ct_ipv4_table[] = {
 713         {
 714                 .ctl_name       = NET_IPV4,
 715                 .procname       = "ipv4",
 716                 .mode           = 0555,
 717                 .child          = ip_ct_netfilter_table,
 718         },
 719         { .ctl_name = 0 }
 720 };
 721
 722 static ctl_table ip_ct_net_table[] = {
 723         {
 724                 .ctl_name       = CTL_NET,
 725                 .procname       = "net",
 726                 .mode           = 0555,
 727                 .child          = ip_ct_ipv4_table,
 728         },
 729         { .ctl_name = 0 }
 730 };
 731
 732 EXPORT_SYMBOL(ip_ct_log_invalid);
 733 #endif /* CONFIG_SYSCTL */
 734
 735 static int init_or_cleanup(int init)
 736 {
 737 #ifdef CONFIG_PROC_FS
 738         struct proc_dir_entry *proc, *proc_exp, *proc_stat;
 739 #endif
 740         int ret = 0;
 741
 742         if (!init) goto cleanup;
 743
 744         ret = ip_conntrack_init();
 745         if (ret < 0)
 746                 goto cleanup_nothing;
 747
 748 #ifdef CONFIG_PROC_FS
 749         proc = proc_net_fops_create("ip_conntrack", 0440, &ct_file_ops);
 750         if (!proc) goto cleanup_init;
 751
 752         proc_exp = proc_net_fops_create("ip_conntrack_expect", 0440,
 753                                         &exp_file_ops);
 754         if (!proc_exp) goto cleanup_proc;
 755
 756         proc_stat = create_proc_entry("ip_conntrack", S_IRUGO, proc_net_stat);
 757         if (!proc_stat)
 758                 goto cleanup_proc_exp;
 759
 760         proc_stat->proc_fops = &ct_cpu_seq_fops;
 761         proc_stat->owner = THIS_MODULE;
 762 #endif
 763
 764         ret = nf_register_hook(&ip_conntrack_defrag_ops);
 765         if (ret < 0) {
 766                 printk("ip_conntrack: can't register pre-routing defrag hook.\n");
 767                 goto cleanup_proc_stat;
 768         }
 769         ret = nf_register_hook(&ip_conntrack_defrag_local_out_ops);
 770         if (ret < 0) {
 771                 printk("ip_conntrack: can't register local_out defrag hook.\n");
 772                 goto cleanup_defragops;
 773         }
 774         ret = nf_register_hook(&ip_conntrack_in_ops);
 775         if (ret < 0) {
 776                 printk("ip_conntrack: can't register pre-routing hook.\n");
 777                 goto cleanup_defraglocalops;
 778         }
 779         ret = nf_register_hook(&ip_conntrack_local_out_ops);
 780         if (ret < 0) {
 781                 printk("ip_conntrack: can't register local out hook.\n");
 782                 goto cleanup_inops;
 783         }
 784         ret = nf_register_hook(&ip_conntrack_out_ops);
 785         if (ret < 0) {
 786                 printk("ip_conntrack: can't register post-routing hook.\n");
 787                 goto cleanup_inandlocalops;
 788         }
 789         ret = nf_register_hook(&ip_conntrack_local_in_ops);
 790         if (ret < 0) {
 791                 printk("ip_conntrack: can't register local in hook.\n");
 792                 goto cleanup_inoutandlocalops;
 793         }
 794 #ifdef CONFIG_SYSCTL
 795         ip_ct_sysctl_header = register_sysctl_table(ip_ct_net_table, 0);
 796         if (ip_ct_sysctl_header == NULL) {
 797                 printk("ip_conntrack: can't register to sysctl.\n");
 798                 goto cleanup;
 799         }
 800 #endif
 801
 802         return ret;
 803
 804  cleanup:
 805 #ifdef CONFIG_SYSCTL
 806         unregister_sysctl_table(ip_ct_sysctl_header);
 807 #endif
 808         nf_unregister_hook(&ip_conntrack_local_in_ops);
 809  cleanup_inoutandlocalops:
 810         nf_unregister_hook(&ip_conntrack_out_ops);
 811  cleanup_inandlocalops:
 812         nf_unregister_hook(&ip_conntrack_local_out_ops);
 813  cleanup_inops:
 814         nf_unregister_hook(&ip_conntrack_in_ops);
 815  cleanup_defraglocalops:
 816         nf_unregister_hook(&ip_conntrack_defrag_local_out_ops);
 817  cleanup_defragops:
 818         /* Frag queues may hold fragments with skb->dst == NULL */
 819         ip_ct_no_defrag = 1;
 820         synchronize_net();
 821         local_bh_disable();
 822         ipfrag_flush();
 823         local_bh_enable();
 824         nf_unregister_hook(&ip_conntrack_defrag_ops);
 825  cleanup_proc_stat:
 826 #ifdef CONFIG_PROC_FS
 827         proc_net_remove("ip_conntrack_stat");
 828 cleanup_proc_exp:
 829         proc_net_remove("ip_conntrack_expect");
 830  cleanup_proc:
 831         proc_net_remove("ip_conntrack");
 832  cleanup_init:
 833 #endif /* CONFIG_PROC_FS */
 834         ip_conntrack_cleanup();
 835  cleanup_nothing:
 836         return ret;
 837 }
 838
 839 /* FIXME: Allow NULL functions and sub in pointers to generic for
 840    them. --RR */
 841 int ip_conntrack_protocol_register(struct ip_conntrack_protocol *proto)
 842 {
 843         int ret = 0;
 844
 845         WRITE_LOCK(&ip_conntrack_lock);
 846         if (ip_ct_protos[proto->proto] != &ip_conntrack_generic_protocol) {
 847                 ret = -EBUSY;
 848                 goto out;
 849         }
 850         ip_ct_protos[proto->proto] = proto;
 851  out:
 852         WRITE_UNLOCK(&ip_conntrack_lock);
 853         return ret;
 854 }
 855
 856 void ip_conntrack_protocol_unregister(struct ip_conntrack_protocol *proto)
 857 {
 858         WRITE_LOCK(&ip_conntrack_lock);
 859         ip_ct_protos[proto->proto] = &ip_conntrack_generic_protocol;
 860         WRITE_UNLOCK(&ip_conntrack_lock);
 861
 862         /* Somebody could be still looking at the proto in bh. */
 863         synchronize_net();
 864
 865         /* Remove all contrack entries for this protocol */
 866         ip_ct_selective_cleanup(kill_proto, &proto->proto);
 867 }
 868
 869 static int __init init(void)
 870 {
 871         return init_or_cleanup(1);
 872 }
 873
 874 static void __exit fini(void)
 875 {
 876         init_or_cleanup(0);
 877 }
 878
 879 module_init(init);
 880 module_exit(fini);
 881
 882 /* Some modules need us, but don't depend directly on any symbol.
 883    They should call this. */
 884 void need_ip_conntrack(void)
 885 {
 886 }
 887
 888 EXPORT_SYMBOL(ip_conntrack_protocol_register);
 889 EXPORT_SYMBOL(ip_conntrack_protocol_unregister);
 890 EXPORT_SYMBOL(invert_tuplepr);
 891 EXPORT_SYMBOL(ip_conntrack_alter_reply);
 892 EXPORT_SYMBOL(ip_conntrack_destroyed);
 893 EXPORT_SYMBOL(need_ip_conntrack);
 894 EXPORT_SYMBOL(ip_conntrack_helper_register);
 895 EXPORT_SYMBOL(ip_conntrack_helper_unregister);
 896 EXPORT_SYMBOL(ip_ct_selective_cleanup);
 897 EXPORT_SYMBOL(ip_ct_refresh_acct);
 898 EXPORT_SYMBOL(ip_ct_protos);
 899 EXPORT_SYMBOL(ip_ct_find_proto);
 900 EXPORT_SYMBOL(ip_ct_find_helper);
 901 EXPORT_SYMBOL(ip_conntrack_expect_alloc);
 902 EXPORT_SYMBOL(ip_conntrack_expect_related);
 903 EXPORT_SYMBOL(ip_conntrack_change_expect);
 904 EXPORT_SYMBOL(ip_conntrack_unexpect_related);
 905 EXPORT_SYMBOL_GPL(ip_conntrack_expect_find_get);
 906 EXPORT_SYMBOL_GPL(ip_conntrack_expect_put);
 907 EXPORT_SYMBOL(ip_conntrack_tuple_taken);
 908 EXPORT_SYMBOL(ip_ct_gather_frags);
 909 EXPORT_SYMBOL(ip_conntrack_htable_size);
 910 EXPORT_SYMBOL(ip_conntrack_expect_list);
 911 EXPORT_SYMBOL(ip_conntrack_lock);
 912 EXPORT_SYMBOL(ip_conntrack_hash);
 913 EXPORT_SYMBOL(ip_conntrack_untracked);
 914 EXPORT_SYMBOL_GPL(ip_conntrack_find_get);
 915 EXPORT_SYMBOL_GPL(ip_conntrack_put);