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