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)
  32 #define ASSERT_WRITE_LOCK(x)
  33
  34 #include <linux/netfilter_ipv4/ip_conntrack.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  37 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  38 #include <linux/netfilter_ipv4/listhelp.h>
  39
  40 #if 0
  41 #define DEBUGP printk
  42 #else
  43 #define DEBUGP(format, args...)
  44 #endif
  45
  46 MODULE_LICENSE("GPL");
  47
  48 extern atomic_t ip_conntrack_count;
  49 DECLARE_PER_CPU(struct ip_conntrack_stat, ip_conntrack_stat);
  50
  51 static int kill_proto(struct ip_conntrack *i, void *data)
  52 {
  53         return (i->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum ==
  54                         *((u_int8_t *) data));
  55 }
  56
  57 #ifdef CONFIG_PROC_FS
  58 static int
  59 print_tuple(struct seq_file *s, const struct ip_conntrack_tuple *tuple,
  60             struct ip_conntrack_protocol *proto)
  61 {
  62         seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  63                    NIPQUAD(tuple->src.ip), NIPQUAD(tuple->dst.ip));
  64         return proto->print_tuple(s, tuple);
  65 }
  66
  67 #ifdef CONFIG_IP_NF_CT_ACCT
  68 static unsigned int
  69 seq_print_counters(struct seq_file *s,
  70                    const struct ip_conntrack_counter *counter)
  71 {
  72         return seq_printf(s, "packets=%llu bytes=%llu ",
  73                           (unsigned long long)counter->packets,
  74                           (unsigned long long)counter->bytes);
  75 }
  76 #else
  77 #define seq_print_counters(x, y)        0
  78 #endif
  79
  80 struct ct_iter_state {
  81         unsigned int bucket;
  82 };
  83
  84 static struct list_head *ct_get_first(struct seq_file *seq)
  85 {
  86         struct ct_iter_state *st = seq->private;
  87
  88         for (st->bucket = 0;
  89              st->bucket < ip_conntrack_htable_size;
  90              st->bucket++) {
  91                 if (!list_empty(&ip_conntrack_hash[st->bucket]))
  92                         return ip_conntrack_hash[st->bucket].next;
  93         }
  94         return NULL;
  95 }
  96
  97 static struct list_head *ct_get_next(struct seq_file *seq, struct list_head *head)
  98 {
  99         struct ct_iter_state *st = seq->private;
 100
 101         head = head->next;
 102         while (head == &ip_conntrack_hash[st->bucket]) {
 103                 if (++st->bucket >= ip_conntrack_htable_size)
 104                         return NULL;
 105                 head = ip_conntrack_hash[st->bucket].next;
 106         }
 107         return head;
 108 }
 109
 110 static struct list_head *ct_get_idx(struct seq_file *seq, loff_t pos)
 111 {
 112         struct list_head *head = ct_get_first(seq);
 113
 114         if (head)
 115                 while (pos && (head = ct_get_next(seq, head)))
 116                         pos--;
 117         return pos ? NULL : head;
 118 }
 119
 120 static void *ct_seq_start(struct seq_file *seq, loff_t *pos)
 121 {
 122         read_lock_bh(&ip_conntrack_lock);
 123         return ct_get_idx(seq, *pos);
 124 }
 125
 126 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
 127 {
 128         (*pos)++;
 129         return ct_get_next(s, v);
 130 }
 131
 132 static void ct_seq_stop(struct seq_file *s, void *v)
 133 {
 134         read_unlock_bh(&ip_conntrack_lock);
 135 }
 136
 137 static int ct_seq_show(struct seq_file *s, void *v)
 138 {
 139         const struct ip_conntrack_tuple_hash *hash = v;
 140         const struct ip_conntrack *conntrack = tuplehash_to_ctrack(hash);
 141         struct ip_conntrack_protocol *proto;
 142
 143         ASSERT_READ_LOCK(&ip_conntrack_lock);
 144         IP_NF_ASSERT(conntrack);
 145
 146         /* we only want to print DIR_ORIGINAL */
 147         if (DIRECTION(hash))
 148                 return 0;
 149
 150         proto = ip_ct_find_proto(conntrack->tuplehash[IP_CT_DIR_ORIGINAL]
 151                                .tuple.dst.protonum);
 152         IP_NF_ASSERT(proto);
 153
 154         if (seq_printf(s, "%-8s %u %ld ",
 155                       proto->name,
 156                       conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum,
 157                       timer_pending(&conntrack->timeout)
 158                       ? (long)(conntrack->timeout.expires - jiffies)/HZ
 159                       : 0) != 0)
 160                 return -ENOSPC;
 161
 162         if (proto->print_conntrack(s, conntrack))
 163                 return -ENOSPC;
 164
 165         if (print_tuple(s, &conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 166                         proto))
 167                 return -ENOSPC;
 168
 169         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_ORIGINAL]))
 170                 return -ENOSPC;
 171
 172         if (!(test_bit(IPS_SEEN_REPLY_BIT, &conntrack->status)))
 173                 if (seq_printf(s, "[UNREPLIED] "))
 174                         return -ENOSPC;
 175
 176         if (print_tuple(s, &conntrack->tuplehash[IP_CT_DIR_REPLY].tuple,
 177                         proto))
 178                 return -ENOSPC;
 179
 180         if (seq_print_counters(s, &conntrack->counters[IP_CT_DIR_REPLY]))
 181                 return -ENOSPC;
 182
 183         if (test_bit(IPS_ASSURED_BIT, &conntrack->status))
 184                 if (seq_printf(s, "[ASSURED] "))
 185                         return -ENOSPC;
 186
 187 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 188         if (seq_printf(s, "mark=%lu ", conntrack->mark))
 189                 return -ENOSPC;
 190 #endif
 191
 192         if (seq_printf(s, "use=%u\n", atomic_read(&conntrack->ct_general.use)))
 193                 return -ENOSPC;
 194
 195         return 0;
 196 }
 197
 198 static struct seq_operations ct_seq_ops = {
 199         .start = ct_seq_start,
 200         .next  = ct_seq_next,
 201         .stop  = ct_seq_stop,
 202         .show  = ct_seq_show
 203 };
 204
 205 static int ct_open(struct inode *inode, struct file *file)
 206 {
 207         struct seq_file *seq;
 208         struct ct_iter_state *st;
 209         int ret;
 210
 211         st = kmalloc(sizeof(struct ct_iter_state), GFP_KERNEL);
 212         if (st == NULL)
 213                 return -ENOMEM;
 214         ret = seq_open(file, &ct_seq_ops);
 215         if (ret)
 216                 goto out_free;
 217         seq          = file->private_data;
 218         seq->private = st;
 219         memset(st, 0, sizeof(struct ct_iter_state));
 220         return ret;
 221 out_free:
 222         kfree(st);
 223         return ret;
 224 }
 225
 226 static struct file_operations ct_file_ops = {
 227         .owner   = THIS_MODULE,
 228         .open    = ct_open,
 229         .read    = seq_read,
 230         .llseek  = seq_lseek,
 231         .release = seq_release_private,
 232 };
 233
 234 /* expects */
 235 static void *exp_seq_start(struct seq_file *s, loff_t *pos)
 236 {
 237         struct list_head *e = &ip_conntrack_expect_list;
 238         loff_t i;
 239
 240         /* strange seq_file api calls stop even if we fail,
 241          * thus we need to grab lock since stop unlocks */
 242         read_lock_bh(&ip_conntrack_lock);
 243
 244         if (list_empty(e))
 245                 return NULL;
 246
 247         for (i = 0; i <= *pos; i++) {
 248                 e = e->next;
 249                 if (e == &ip_conntrack_expect_list)
 250                         return NULL;
 251         }
 252         return e;
 253 }
 254
 255 static void *exp_seq_next(struct seq_file *s, void *v, loff_t *pos)
 256 {
 257         struct list_head *e = v;
 258
 259         ++*pos;
 260         e = e->next;
 261
 262         if (e == &ip_conntrack_expect_list)
 263                 return NULL;
 264
 265         return e;
 266 }
 267
 268 static void exp_seq_stop(struct seq_file *s, void *v)
 269 {
 270         read_unlock_bh(&ip_conntrack_lock);
 271 }
 272
 273 static int exp_seq_show(struct seq_file *s, void *v)
 274 {
 275         struct ip_conntrack_expect *expect = v;
 276
 277         if (expect->timeout.function)
 278                 seq_printf(s, "%ld ", timer_pending(&expect->timeout)
 279                            ? (long)(expect->timeout.expires - jiffies)/HZ : 0);
 280         else
 281                 seq_printf(s, "- ");
 282
 283         seq_printf(s, "proto=%u ", expect->tuple.dst.protonum);
 284
 285         print_tuple(s, &expect->tuple,
 286                     ip_ct_find_proto(expect->tuple.dst.protonum));
 287         return seq_putc(s, '\n');
 288 }
 289
 290 static struct seq_operations exp_seq_ops = {
 291         .start = exp_seq_start,
 292         .next = exp_seq_next,
 293         .stop = exp_seq_stop,
 294         .show = exp_seq_show
 295 };
 296
 297 static int exp_open(struct inode *inode, struct file *file)
 298 {
 299         return seq_open(file, &exp_seq_ops);
 300 }
 301
 302 static struct file_operations exp_file_ops = {
 303         .owner   = THIS_MODULE,
 304         .open    = exp_open,
 305         .read    = seq_read,
 306         .llseek  = seq_lseek,
 307         .release = seq_release
 308 };
 309
 310 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 311 {
 312         int cpu;
 313
 314         if (*pos == 0)
 315                 return SEQ_START_TOKEN;
 316
 317         for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
 318                 if (!cpu_possible(cpu))
 319                         continue;
 320                 *pos = cpu+1;
 321                 return &per_cpu(ip_conntrack_stat, cpu);
 322         }
 323
 324         return NULL;
 325 }
 326
 327 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 328 {
 329         int cpu;
 330
 331         for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
 332                 if (!cpu_possible(cpu))
 333                         continue;
 334                 *pos = cpu+1;
 335                 return &per_cpu(ip_conntrack_stat, cpu);
 336         }
 337
 338         return NULL;
 339 }
 340
 341 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 342 {
 343 }
 344
 345 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 346 {
 347         unsigned int nr_conntracks = atomic_read(&ip_conntrack_count);
 348         struct ip_conntrack_stat *st = v;
 349
 350         if (v == SEQ_START_TOKEN) {
 351                 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");
 352                 return 0;
 353         }
 354
 355         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 356                         "%08x %08x %08x %08x %08x  %08x %08x %08x \n",
 357                    nr_conntracks,
 358                    st->searched,
 359                    st->found,
 360                    st->new,
 361                    st->invalid,
 362                    st->ignore,
 363                    st->delete,
 364                    st->delete_list,
 365                    st->insert,
 366                    st->insert_failed,
 367                    st->drop,
 368                    st->early_drop,
 369                    st->error,
 370
 371                    st->expect_new,
 372                    st->expect_create,
 373                    st->expect_delete
 374                 );
 375         return 0;
 376 }
 377
 378 static struct seq_operations ct_cpu_seq_ops = {
 379         .start  = ct_cpu_seq_start,
 380         .next   = ct_cpu_seq_next,
 381         .stop   = ct_cpu_seq_stop,
 382         .show   = ct_cpu_seq_show,
 383 };
 384
 385 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 386 {
 387         return seq_open(file, &ct_cpu_seq_ops);
 388 }
 389
 390 static struct file_operations ct_cpu_seq_fops = {
 391         .owner   = THIS_MODULE,
 392         .open    = ct_cpu_seq_open,
 393         .read    = seq_read,
 394         .llseek  = seq_lseek,
 395         .release = seq_release_private,
 396 };
 397 #endif
 398
 399 static unsigned int ip_confirm(unsigned int hooknum,
 400                                struct sk_buff **pskb,
 401                                const struct net_device *in,
 402                                const struct net_device *out,
 403                                int (*okfn)(struct sk_buff *))
 404 {
 405         /* We've seen it coming out the other side: confirm it */
 406         return ip_conntrack_confirm(pskb);
 407 }
 408
 409 static unsigned int ip_conntrack_help(unsigned int hooknum,
 410                                       struct sk_buff **pskb,
 411                                       const struct net_device *in,
 412                                       const struct net_device *out,
 413                                       int (*okfn)(struct sk_buff *))
 414 {
 415         struct ip_conntrack *ct;
 416         enum ip_conntrack_info ctinfo;
 417
 418         /* This is where we call the helper: as the packet goes out. */
 419         ct = ip_conntrack_get(*pskb, &ctinfo);
 420         if (ct && ct->helper) {
 421                 unsigned int ret;
 422                 ret = ct->helper->help(pskb, ct, ctinfo);
 423                 if (ret != NF_ACCEPT)
 424                         return ret;
 425         }
 426         return NF_ACCEPT;
 427 }
 428
 429 static unsigned int ip_conntrack_defrag(unsigned int hooknum,
 430                                         struct sk_buff **pskb,
 431                                         const struct net_device *in,
 432                                         const struct net_device *out,
 433                                         int (*okfn)(struct sk_buff *))
 434 {
 435 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
 436         /* Previously seen (loopback)?  Ignore.  Do this before
 437            fragment check. */
 438         if ((*pskb)->nfct)
 439                 return NF_ACCEPT;
 440 #endif
 441
 442         /* Gather fragments. */
 443         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 444                 *pskb = ip_ct_gather_frags(*pskb,
 445                                            hooknum == NF_IP_PRE_ROUTING ?
 446                                            IP_DEFRAG_CONNTRACK_IN :
 447                                            IP_DEFRAG_CONNTRACK_OUT);
 448                 if (!*pskb)
 449                         return NF_STOLEN;
 450         }
 451         return NF_ACCEPT;
 452 }
 453
 454 static unsigned int ip_refrag(unsigned int hooknum,
 455                               struct sk_buff **pskb,
 456                               const struct net_device *in,
 457                               const struct net_device *out,
 458                               int (*okfn)(struct sk_buff *))
 459 {
 460         struct rtable *rt = (struct rtable *)(*pskb)->dst;
 461
 462         /* We've seen it coming out the other side: confirm */
 463         if (ip_confirm(hooknum, pskb, in, out, okfn) != NF_ACCEPT)
 464                 return NF_DROP;
 465
 466         /* Local packets are never produced too large for their
 467            interface.  We degfragment them at LOCAL_OUT, however,
 468            so we have to refragment them here. */
 469         if ((*pskb)->len > dst_mtu(&rt->u.dst) &&
 470             !skb_shinfo(*pskb)->tso_size) {
 471                 /* No hook can be after us, so this should be OK. */
 472                 ip_fragment(*pskb, okfn);
 473                 return NF_STOLEN;
 474         }
 475         return NF_ACCEPT;
 476 }
 477
 478 static unsigned int ip_conntrack_local(unsigned int hooknum,
 479                                        struct sk_buff **pskb,
 480                                        const struct net_device *in,
 481                                        const struct net_device *out,
 482                                        int (*okfn)(struct sk_buff *))
 483 {
 484         /* root is playing with raw sockets. */
 485         if ((*pskb)->len < sizeof(struct iphdr)
 486             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 487                 if (net_ratelimit())
 488                         printk("ipt_hook: happy cracking.\n");
 489                 return NF_ACCEPT;
 490         }
 491         return ip_conntrack_in(hooknum, pskb, in, out, okfn);
 492 }
 493
 494 /* Connection tracking may drop packets, but never alters them, so
 495    make it the first hook. */
 496 static struct nf_hook_ops ip_conntrack_defrag_ops = {
 497         .hook           = ip_conntrack_defrag,
 498         .owner          = THIS_MODULE,
 499         .pf             = PF_INET,
 500         .hooknum        = NF_IP_PRE_ROUTING,
 501         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 502 };
 503
 504 static struct nf_hook_ops ip_conntrack_in_ops = {
 505         .hook           = ip_conntrack_in,
 506         .owner          = THIS_MODULE,
 507         .pf             = PF_INET,
 508         .hooknum        = NF_IP_PRE_ROUTING,
 509         .priority       = NF_IP_PRI_CONNTRACK,
 510 };
 511
 512 static struct nf_hook_ops ip_conntrack_defrag_local_out_ops = {
 513         .hook           = ip_conntrack_defrag,
 514         .owner          = THIS_MODULE,
 515         .pf             = PF_INET,
 516         .hooknum        = NF_IP_LOCAL_OUT,
 517         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 518 };
 519
 520 static struct nf_hook_ops ip_conntrack_local_out_ops = {
 521         .hook           = ip_conntrack_local,
 522         .owner          = THIS_MODULE,
 523         .pf             = PF_INET,
 524         .hooknum        = NF_IP_LOCAL_OUT,
 525         .priority       = NF_IP_PRI_CONNTRACK,
 526 };
 527
 528 /* helpers */
 529 static struct nf_hook_ops ip_conntrack_helper_out_ops = {
 530         .hook           = ip_conntrack_help,
 531         .owner          = THIS_MODULE,
 532         .pf             = PF_INET,
 533         .hooknum        = NF_IP_POST_ROUTING,
 534         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 535 };
 536
 537 static struct nf_hook_ops ip_conntrack_helper_in_ops = {
 538         .hook           = ip_conntrack_help,
 539         .owner          = THIS_MODULE,
 540         .pf             = PF_INET,
 541         .hooknum        = NF_IP_LOCAL_IN,
 542         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 543 };
 544
 545 /* Refragmenter; last chance. */
 546 static struct nf_hook_ops ip_conntrack_out_ops = {
 547         .hook           = ip_refrag,
 548         .owner          = THIS_MODULE,
 549         .pf             = PF_INET,
 550         .hooknum        = NF_IP_POST_ROUTING,
 551         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 552 };
 553
 554 static struct nf_hook_ops ip_conntrack_local_in_ops = {
 555         .hook           = ip_confirm,
 556         .owner          = THIS_MODULE,
 557         .pf             = PF_INET,
 558         .hooknum        = NF_IP_LOCAL_IN,
 559         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 560 };
 561
 562 /* Sysctl support */
 563
 564 #ifdef CONFIG_SYSCTL
 565
 566 /* From ip_conntrack_core.c */
 567 extern int ip_conntrack_max;
 568 extern unsigned int ip_conntrack_htable_size;
 569
 570 /* From ip_conntrack_proto_tcp.c */
 571 extern unsigned long ip_ct_tcp_timeout_syn_sent;
 572 extern unsigned long ip_ct_tcp_timeout_syn_recv;
 573 extern unsigned long ip_ct_tcp_timeout_established;
 574 extern unsigned long ip_ct_tcp_timeout_fin_wait;
 575 extern unsigned long ip_ct_tcp_timeout_close_wait;
 576 extern unsigned long ip_ct_tcp_timeout_last_ack;
 577 extern unsigned long ip_ct_tcp_timeout_time_wait;
 578 extern unsigned long ip_ct_tcp_timeout_close;
 579 extern unsigned long ip_ct_tcp_timeout_max_retrans;
 580 extern int ip_ct_tcp_loose;
 581 extern int ip_ct_tcp_be_liberal;
 582 extern int ip_ct_tcp_max_retrans;
 583
 584 /* From ip_conntrack_proto_udp.c */
 585 extern unsigned long ip_ct_udp_timeout;
 586 extern unsigned long ip_ct_udp_timeout_stream;
 587
 588 /* From ip_conntrack_proto_icmp.c */
 589 extern unsigned long ip_ct_icmp_timeout;
 590
 591 /* From ip_conntrack_proto_icmp.c */
 592 extern unsigned long ip_ct_generic_timeout;
 593
 594 /* Log invalid packets of a given protocol */
 595 static int log_invalid_proto_min = 0;
 596 static int log_invalid_proto_max = 255;
 597
 598 static struct ctl_table_header *ip_ct_sysctl_header;
 599
 600 static ctl_table ip_ct_sysctl_table[] = {
 601         {
 602                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 603                 .procname       = "ip_conntrack_max",
 604                 .data           = &ip_conntrack_max,
 605                 .maxlen         = sizeof(int),
 606                 .mode           = 0644,
 607                 .proc_handler   = &proc_dointvec,
 608         },
 609         {
 610                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 611                 .procname       = "ip_conntrack_count",
 612                 .data           = &ip_conntrack_count,
 613                 .maxlen         = sizeof(int),
 614                 .mode           = 0444,
 615                 .proc_handler   = &proc_dointvec,
 616         },
 617         {
 618                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 619                 .procname       = "ip_conntrack_buckets",
 620                 .data           = &ip_conntrack_htable_size,
 621                 .maxlen         = sizeof(unsigned int),
 622                 .mode           = 0444,
 623                 .proc_handler   = &proc_dointvec,
 624         },
 625         {
 626                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_SENT,
 627                 .procname       = "ip_conntrack_tcp_timeout_syn_sent",
 628                 .data           = &ip_ct_tcp_timeout_syn_sent,
 629                 .maxlen         = sizeof(unsigned int),
 630                 .mode           = 0644,
 631                 .proc_handler   = &proc_dointvec_jiffies,
 632         },
 633         {
 634                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_SYN_RECV,
 635                 .procname       = "ip_conntrack_tcp_timeout_syn_recv",
 636                 .data           = &ip_ct_tcp_timeout_syn_recv,
 637                 .maxlen         = sizeof(unsigned int),
 638                 .mode           = 0644,
 639                 .proc_handler   = &proc_dointvec_jiffies,
 640         },
 641         {
 642                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_ESTABLISHED,
 643                 .procname       = "ip_conntrack_tcp_timeout_established",
 644                 .data           = &ip_ct_tcp_timeout_established,
 645                 .maxlen         = sizeof(unsigned int),
 646                 .mode           = 0644,
 647                 .proc_handler   = &proc_dointvec_jiffies,
 648         },
 649         {
 650                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_FIN_WAIT,
 651                 .procname       = "ip_conntrack_tcp_timeout_fin_wait",
 652                 .data           = &ip_ct_tcp_timeout_fin_wait,
 653                 .maxlen         = sizeof(unsigned int),
 654                 .mode           = 0644,
 655                 .proc_handler   = &proc_dointvec_jiffies,
 656         },
 657         {
 658                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE_WAIT,
 659                 .procname       = "ip_conntrack_tcp_timeout_close_wait",
 660                 .data           = &ip_ct_tcp_timeout_close_wait,
 661                 .maxlen         = sizeof(unsigned int),
 662                 .mode           = 0644,
 663                 .proc_handler   = &proc_dointvec_jiffies,
 664         },
 665         {
 666                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_LAST_ACK,
 667                 .procname       = "ip_conntrack_tcp_timeout_last_ack",
 668                 .data           = &ip_ct_tcp_timeout_last_ack,
 669                 .maxlen         = sizeof(unsigned int),
 670                 .mode           = 0644,
 671                 .proc_handler   = &proc_dointvec_jiffies,
 672         },
 673         {
 674                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_TIME_WAIT,
 675                 .procname       = "ip_conntrack_tcp_timeout_time_wait",
 676                 .data           = &ip_ct_tcp_timeout_time_wait,
 677                 .maxlen         = sizeof(unsigned int),
 678                 .mode           = 0644,
 679                 .proc_handler   = &proc_dointvec_jiffies,
 680         },
 681         {
 682                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_CLOSE,
 683                 .procname       = "ip_conntrack_tcp_timeout_close",
 684                 .data           = &ip_ct_tcp_timeout_close,
 685                 .maxlen         = sizeof(unsigned int),
 686                 .mode           = 0644,
 687                 .proc_handler   = &proc_dointvec_jiffies,
 688         },
 689         {
 690                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT,
 691                 .procname       = "ip_conntrack_udp_timeout",
 692                 .data           = &ip_ct_udp_timeout,
 693                 .maxlen         = sizeof(unsigned int),
 694                 .mode           = 0644,
 695                 .proc_handler   = &proc_dointvec_jiffies,
 696         },
 697         {
 698                 .ctl_name       = NET_IPV4_NF_CONNTRACK_UDP_TIMEOUT_STREAM,
 699                 .procname       = "ip_conntrack_udp_timeout_stream",
 700                 .data           = &ip_ct_udp_timeout_stream,
 701                 .maxlen         = sizeof(unsigned int),
 702                 .mode           = 0644,
 703                 .proc_handler   = &proc_dointvec_jiffies,
 704         },
 705         {
 706                 .ctl_name       = NET_IPV4_NF_CONNTRACK_ICMP_TIMEOUT,
 707                 .procname       = "ip_conntrack_icmp_timeout",
 708                 .data           = &ip_ct_icmp_timeout,
 709                 .maxlen         = sizeof(unsigned int),
 710                 .mode           = 0644,
 711                 .proc_handler   = &proc_dointvec_jiffies,
 712         },
 713         {
 714                 .ctl_name       = NET_IPV4_NF_CONNTRACK_GENERIC_TIMEOUT,
 715                 .procname       = "ip_conntrack_generic_timeout",
 716                 .data           = &ip_ct_generic_timeout,
 717                 .maxlen         = sizeof(unsigned int),
 718                 .mode           = 0644,
 719                 .proc_handler   = &proc_dointvec_jiffies,
 720         },
 721         {
 722                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 723                 .procname       = "ip_conntrack_log_invalid",
 724                 .data           = &ip_ct_log_invalid,
 725                 .maxlen         = sizeof(unsigned int),
 726                 .mode           = 0644,
 727                 .proc_handler   = &proc_dointvec_minmax,
 728                 .strategy       = &sysctl_intvec,
 729                 .extra1         = &log_invalid_proto_min,
 730                 .extra2         = &log_invalid_proto_max,
 731         },
 732         {
 733                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_TIMEOUT_MAX_RETRANS,
 734                 .procname       = "ip_conntrack_tcp_timeout_max_retrans",
 735                 .data           = &ip_ct_tcp_timeout_max_retrans,
 736                 .maxlen         = sizeof(unsigned int),
 737                 .mode           = 0644,
 738                 .proc_handler   = &proc_dointvec_jiffies,
 739         },
 740         {
 741                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_LOOSE,
 742                 .procname       = "ip_conntrack_tcp_loose",
 743                 .data           = &ip_ct_tcp_loose,
 744                 .maxlen         = sizeof(unsigned int),
 745                 .mode           = 0644,
 746                 .proc_handler   = &proc_dointvec,
 747         },
 748         {
 749                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_BE_LIBERAL,
 750                 .procname       = "ip_conntrack_tcp_be_liberal",
 751                 .data           = &ip_ct_tcp_be_liberal,
 752                 .maxlen         = sizeof(unsigned int),
 753                 .mode           = 0644,
 754                 .proc_handler   = &proc_dointvec,
 755         },
 756         {
 757                 .ctl_name       = NET_IPV4_NF_CONNTRACK_TCP_MAX_RETRANS,
 758                 .procname       = "ip_conntrack_tcp_max_retrans",
 759                 .data           = &ip_ct_tcp_max_retrans,
 760                 .maxlen         = sizeof(unsigned int),
 761                 .mode           = 0644,
 762                 .proc_handler   = &proc_dointvec,
 763         },
 764         { .ctl_name = 0 }
 765 };
 766
 767 #define NET_IP_CONNTRACK_MAX 2089
 768
 769 static ctl_table ip_ct_netfilter_table[] = {
 770         {
 771                 .ctl_name       = NET_IPV4_NETFILTER,
 772                 .procname       = "netfilter",
 773                 .mode           = 0555,
 774                 .child          = ip_ct_sysctl_table,
 775         },
 776         {
 777                 .ctl_name       = NET_IP_CONNTRACK_MAX,
 778                 .procname       = "ip_conntrack_max",
 779                 .data           = &ip_conntrack_max,
 780                 .maxlen         = sizeof(int),
 781                 .mode           = 0644,
 782                 .proc_handler   = &proc_dointvec
 783         },
 784         { .ctl_name = 0 }
 785 };
 786
 787 static ctl_table ip_ct_ipv4_table[] = {
 788         {
 789                 .ctl_name       = NET_IPV4,
 790                 .procname       = "ipv4",
 791                 .mode           = 0555,
 792                 .child          = ip_ct_netfilter_table,
 793         },
 794         { .ctl_name = 0 }
 795 };
 796
 797 static ctl_table ip_ct_net_table[] = {
 798         {
 799                 .ctl_name       = CTL_NET,
 800                 .procname       = "net",
 801                 .mode           = 0555,
 802                 .child          = ip_ct_ipv4_table,
 803         },
 804         { .ctl_name = 0 }
 805 };
 806
 807 EXPORT_SYMBOL(ip_ct_log_invalid);
 808 #endif /* CONFIG_SYSCTL */
 809
 810 static int init_or_cleanup(int init)
 811 {
 812 #ifdef CONFIG_PROC_FS
 813         struct proc_dir_entry *proc, *proc_exp, *proc_stat;
 814 #endif
 815         int ret = 0;
 816
 817         if (!init) goto cleanup;
 818
 819         ret = ip_conntrack_init();
 820         if (ret < 0)
 821                 goto cleanup_nothing;
 822
 823 #ifdef CONFIG_PROC_FS
 824         ret = -ENOMEM;
 825         proc = proc_net_fops_create("ip_conntrack", 0440, &ct_file_ops);
 826         if (!proc) goto cleanup_init;
 827
 828         proc_exp = proc_net_fops_create("ip_conntrack_expect", 0440,
 829                                         &exp_file_ops);
 830         if (!proc_exp) goto cleanup_proc;
 831
 832         proc_stat = create_proc_entry("ip_conntrack", S_IRUGO, proc_net_stat);
 833         if (!proc_stat)
 834                 goto cleanup_proc_exp;
 835
 836         proc_stat->proc_fops = &ct_cpu_seq_fops;
 837         proc_stat->owner = THIS_MODULE;
 838 #endif
 839
 840         ret = nf_register_hook(&ip_conntrack_defrag_ops);
 841         if (ret < 0) {
 842                 printk("ip_conntrack: can't register pre-routing defrag hook.\n");
 843                 goto cleanup_proc_stat;
 844         }
 845         ret = nf_register_hook(&ip_conntrack_defrag_local_out_ops);
 846         if (ret < 0) {
 847                 printk("ip_conntrack: can't register local_out defrag hook.\n");
 848                 goto cleanup_defragops;
 849         }
 850         ret = nf_register_hook(&ip_conntrack_in_ops);
 851         if (ret < 0) {
 852                 printk("ip_conntrack: can't register pre-routing hook.\n");
 853                 goto cleanup_defraglocalops;
 854         }
 855         ret = nf_register_hook(&ip_conntrack_local_out_ops);
 856         if (ret < 0) {
 857                 printk("ip_conntrack: can't register local out hook.\n");
 858                 goto cleanup_inops;
 859         }
 860         ret = nf_register_hook(&ip_conntrack_helper_in_ops);
 861         if (ret < 0) {
 862                 printk("ip_conntrack: can't register local in helper hook.\n");
 863                 goto cleanup_inandlocalops;
 864         }
 865         ret = nf_register_hook(&ip_conntrack_helper_out_ops);
 866         if (ret < 0) {
 867                 printk("ip_conntrack: can't register postrouting helper hook.\n");
 868                 goto cleanup_helperinops;
 869         }
 870         ret = nf_register_hook(&ip_conntrack_out_ops);
 871         if (ret < 0) {
 872                 printk("ip_conntrack: can't register post-routing hook.\n");
 873                 goto cleanup_helperoutops;
 874         }
 875         ret = nf_register_hook(&ip_conntrack_local_in_ops);
 876         if (ret < 0) {
 877                 printk("ip_conntrack: can't register local in hook.\n");
 878                 goto cleanup_inoutandlocalops;
 879         }
 880 #ifdef CONFIG_SYSCTL
 881         ip_ct_sysctl_header = register_sysctl_table(ip_ct_net_table, 0);
 882         if (ip_ct_sysctl_header == NULL) {
 883                 printk("ip_conntrack: can't register to sysctl.\n");
 884                 ret = -ENOMEM;
 885                 goto cleanup_localinops;
 886         }
 887 #endif
 888
 889         return ret;
 890
 891  cleanup:
 892 #ifdef CONFIG_SYSCTL
 893         unregister_sysctl_table(ip_ct_sysctl_header);
 894  cleanup_localinops:
 895 #endif
 896         nf_unregister_hook(&ip_conntrack_local_in_ops);
 897  cleanup_inoutandlocalops:
 898         nf_unregister_hook(&ip_conntrack_out_ops);
 899  cleanup_helperoutops:
 900         nf_unregister_hook(&ip_conntrack_helper_out_ops);
 901  cleanup_helperinops:
 902         nf_unregister_hook(&ip_conntrack_helper_in_ops);
 903  cleanup_inandlocalops:
 904         nf_unregister_hook(&ip_conntrack_local_out_ops);
 905  cleanup_inops:
 906         nf_unregister_hook(&ip_conntrack_in_ops);
 907  cleanup_defraglocalops:
 908         nf_unregister_hook(&ip_conntrack_defrag_local_out_ops);
 909  cleanup_defragops:
 910         nf_unregister_hook(&ip_conntrack_defrag_ops);
 911  cleanup_proc_stat:
 912 #ifdef CONFIG_PROC_FS
 913         remove_proc_entry("ip_conntrack", proc_net_stat);
 914  cleanup_proc_exp:
 915         proc_net_remove("ip_conntrack_expect");
 916  cleanup_proc:
 917         proc_net_remove("ip_conntrack");
 918  cleanup_init:
 919 #endif /* CONFIG_PROC_FS */
 920         ip_conntrack_cleanup();
 921  cleanup_nothing:
 922         return ret;
 923 }
 924
 925 /* FIXME: Allow NULL functions and sub in pointers to generic for
 926    them. --RR */
 927 int ip_conntrack_protocol_register(struct ip_conntrack_protocol *proto)
 928 {
 929         int ret = 0;
 930
 931         write_lock_bh(&ip_conntrack_lock);
 932         if (ip_ct_protos[proto->proto] != &ip_conntrack_generic_protocol) {
 933                 ret = -EBUSY;
 934                 goto out;
 935         }
 936         ip_ct_protos[proto->proto] = proto;
 937  out:
 938         write_unlock_bh(&ip_conntrack_lock);
 939         return ret;
 940 }
 941
 942 void ip_conntrack_protocol_unregister(struct ip_conntrack_protocol *proto)
 943 {
 944         write_lock_bh(&ip_conntrack_lock);
 945         ip_ct_protos[proto->proto] = &ip_conntrack_generic_protocol;
 946         write_unlock_bh(&ip_conntrack_lock);
 947
 948         /* Somebody could be still looking at the proto in bh. */
 949         synchronize_net();
 950
 951         /* Remove all contrack entries for this protocol */
 952         ip_ct_iterate_cleanup(kill_proto, &proto->proto);
 953 }
 954
 955 static int __init init(void)
 956 {
 957         return init_or_cleanup(1);
 958 }
 959
 960 static void __exit fini(void)
 961 {
 962         init_or_cleanup(0);
 963 }
 964
 965 module_init(init);
 966 module_exit(fini);
 967
 968 /* Some modules need us, but don't depend directly on any symbol.
 969    They should call this. */
 970 void need_ip_conntrack(void)
 971 {
 972 }
 973
 974 EXPORT_SYMBOL(ip_conntrack_protocol_register);
 975 EXPORT_SYMBOL(ip_conntrack_protocol_unregister);
 976 EXPORT_SYMBOL(ip_ct_get_tuple);
 977 EXPORT_SYMBOL(invert_tuplepr);
 978 EXPORT_SYMBOL(ip_conntrack_alter_reply);
 979 EXPORT_SYMBOL(ip_conntrack_destroyed);
 980 EXPORT_SYMBOL(need_ip_conntrack);
 981 EXPORT_SYMBOL(ip_conntrack_helper_register);
 982 EXPORT_SYMBOL(ip_conntrack_helper_unregister);
 983 EXPORT_SYMBOL(ip_ct_iterate_cleanup);
 984 EXPORT_SYMBOL(ip_ct_refresh_acct);
 985 EXPORT_SYMBOL(ip_ct_protos);
 986 EXPORT_SYMBOL(ip_ct_find_proto);
 987 EXPORT_SYMBOL(ip_conntrack_expect_alloc);
 988 EXPORT_SYMBOL(ip_conntrack_expect_free);
 989 EXPORT_SYMBOL(ip_conntrack_expect_related);
 990 EXPORT_SYMBOL(ip_conntrack_unexpect_related);
 991 EXPORT_SYMBOL(ip_conntrack_tuple_taken);
 992 EXPORT_SYMBOL(ip_ct_gather_frags);
 993 EXPORT_SYMBOL(ip_conntrack_htable_size);
 994 EXPORT_SYMBOL(ip_conntrack_lock);
 995 EXPORT_SYMBOL(ip_conntrack_hash);
 996 EXPORT_SYMBOL(ip_conntrack_untracked);
 997 EXPORT_SYMBOL_GPL(ip_conntrack_find_get);
 998 EXPORT_SYMBOL_GPL(ip_conntrack_put);
 999 #ifdef CONFIG_IP_NF_NAT_NEEDED
1000 EXPORT_SYMBOL(ip_conntrack_tcp_update);
1001 #endif