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