net/ipv4/netfilter/ip_conntrack_standalone.c

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