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