net/netfilter/nf_conntrack_standalone.c

   1 /* (C) 1999-2001 Paul `Rusty' Russell
   2  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
   3  * (C) 2005-2012 Patrick McHardy <kaber@trash.net>
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License version 2 as
   7  * published by the Free Software Foundation.
   8  */
   9
  10 #include <linux/types.h>
  11 #include <linux/netfilter.h>
  12 #include <linux/slab.h>
  13 #include <linux/module.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/proc_fs.h>
  16 #include <linux/seq_file.h>
  17 #include <linux/percpu.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/security.h>
  20 #include <net/net_namespace.h>
  21 #ifdef CONFIG_SYSCTL
  22 #include <linux/sysctl.h>
  23 #endif
  24
  25 #include <net/netfilter/nf_conntrack.h>
  26 #include <net/netfilter/nf_conntrack_core.h>
  27 #include <net/netfilter/nf_conntrack_l3proto.h>
  28 #include <net/netfilter/nf_conntrack_l4proto.h>
  29 #include <net/netfilter/nf_conntrack_expect.h>
  30 #include <net/netfilter/nf_conntrack_helper.h>
  31 #include <net/netfilter/nf_conntrack_acct.h>
  32 #include <net/netfilter/nf_conntrack_zones.h>
  33 #include <net/netfilter/nf_conntrack_timestamp.h>
  34 #include <linux/rculist_nulls.h>
  35
  36 MODULE_LICENSE("GPL");
  37
  38 #ifdef CONFIG_NF_CONNTRACK_PROCFS
  39 void
  40 print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
  41             const struct nf_conntrack_l3proto *l3proto,
  42             const struct nf_conntrack_l4proto *l4proto)
  43 {
  44         l3proto->print_tuple(s, tuple);
  45         l4proto->print_tuple(s, tuple);
  46 }
  47 EXPORT_SYMBOL_GPL(print_tuple);
  48
  49 struct ct_iter_state {
  50         struct seq_net_private p;
  51         unsigned int bucket;
  52         u_int64_t time_now;
  53 };
  54
  55 static struct hlist_nulls_node *ct_get_first(struct seq_file *seq)
  56 {
  57         struct net *net = seq_file_net(seq);
  58         struct ct_iter_state *st = seq->private;
  59         struct hlist_nulls_node *n;
  60
  61         for (st->bucket = 0;
  62              st->bucket < net->ct.htable_size;
  63              st->bucket++) {
  64                 n = rcu_dereference(hlist_nulls_first_rcu(&net->ct.hash[st->bucket]));
  65                 if (!is_a_nulls(n))
  66                         return n;
  67         }
  68         return NULL;
  69 }
  70
  71 static struct hlist_nulls_node *ct_get_next(struct seq_file *seq,
  72                                       struct hlist_nulls_node *head)
  73 {
  74         struct net *net = seq_file_net(seq);
  75         struct ct_iter_state *st = seq->private;
  76
  77         head = rcu_dereference(hlist_nulls_next_rcu(head));
  78         while (is_a_nulls(head)) {
  79                 if (likely(get_nulls_value(head) == st->bucket)) {
  80                         if (++st->bucket >= net->ct.htable_size)
  81                                 return NULL;
  82                 }
  83                 head = rcu_dereference(
  84                                 hlist_nulls_first_rcu(
  85                                         &net->ct.hash[st->bucket]));
  86         }
  87         return head;
  88 }
  89
  90 static struct hlist_nulls_node *ct_get_idx(struct seq_file *seq, loff_t pos)
  91 {
  92         struct hlist_nulls_node *head = ct_get_first(seq);
  93
  94         if (head)
  95                 while (pos && (head = ct_get_next(seq, head)))
  96                         pos--;
  97         return pos ? NULL : head;
  98 }
  99
 100 static void *ct_seq_start(struct seq_file *seq, loff_t *pos)
 101         __acquires(RCU)
 102 {
 103         struct ct_iter_state *st = seq->private;
 104
 105         st->time_now = ktime_get_real_ns();
 106         rcu_read_lock();
 107         return ct_get_idx(seq, *pos);
 108 }
 109
 110 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
 111 {
 112         (*pos)++;
 113         return ct_get_next(s, v);
 114 }
 115
 116 static void ct_seq_stop(struct seq_file *s, void *v)
 117         __releases(RCU)
 118 {
 119         rcu_read_unlock();
 120 }
 121
 122 #ifdef CONFIG_NF_CONNTRACK_SECMARK
 123 static void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
 124 {
 125         int ret;
 126         u32 len;
 127         char *secctx;
 128
 129         ret = security_secid_to_secctx(ct->secmark, &secctx, &len);
 130         if (ret)
 131                 return;
 132
 133         seq_printf(s, "secctx=%s ", secctx);
 134
 135         security_release_secctx(secctx, len);
 136 }
 137 #else
 138 static inline void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
 139 {
 140 }
 141 #endif
 142
 143 #ifdef CONFIG_NF_CONNTRACK_ZONES
 144 static void ct_show_zone(struct seq_file *s, const struct nf_conn *ct)
 145 {
 146         seq_printf(s, "zone=%u ", nf_ct_zone(ct)->id);
 147 }
 148 #else
 149 static inline void ct_show_zone(struct seq_file *s, const struct nf_conn *ct)
 150 {
 151 }
 152 #endif
 153
 154 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
 155 static void ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 156 {
 157         struct ct_iter_state *st = s->private;
 158         struct nf_conn_tstamp *tstamp;
 159         s64 delta_time;
 160
 161         tstamp = nf_conn_tstamp_find(ct);
 162         if (tstamp) {
 163                 delta_time = st->time_now - tstamp->start;
 164                 if (delta_time > 0)
 165                         delta_time = div_s64(delta_time, NSEC_PER_SEC);
 166                 else
 167                         delta_time = 0;
 168
 169                 seq_printf(s, "delta-time=%llu ",
 170                            (unsigned long long)delta_time);
 171         }
 172         return;
 173 }
 174 #else
 175 static inline void
 176 ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 177 {
 178 }
 179 #endif
 180
 181 /* return 0 on success, 1 in case of error */
 182 static int ct_seq_show(struct seq_file *s, void *v)
 183 {
 184         struct nf_conntrack_tuple_hash *hash = v;
 185         struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash);
 186         const struct nf_conntrack_l3proto *l3proto;
 187         const struct nf_conntrack_l4proto *l4proto;
 188         int ret = 0;
 189
 190         NF_CT_ASSERT(ct);
 191         if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
 192                 return 0;
 193
 194         /* we only want to print DIR_ORIGINAL */
 195         if (NF_CT_DIRECTION(hash))
 196                 goto release;
 197
 198         l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
 199         NF_CT_ASSERT(l3proto);
 200         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
 201         NF_CT_ASSERT(l4proto);
 202
 203         ret = -ENOSPC;
 204         seq_printf(s, "%-8s %u %-8s %u %ld ",
 205                    l3proto->name, nf_ct_l3num(ct),
 206                    l4proto->name, nf_ct_protonum(ct),
 207                    timer_pending(&ct->timeout)
 208                    ? (long)(ct->timeout.expires - jiffies)/HZ : 0);
 209
 210         if (l4proto->print_conntrack)
 211                 l4proto->print_conntrack(s, ct);
 212
 213         print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 214                     l3proto, l4proto);
 215
 216         if (seq_has_overflowed(s))
 217                 goto release;
 218
 219         if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL))
 220                 goto release;
 221
 222         if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status)))
 223                 seq_printf(s, "[UNREPLIED] ");
 224
 225         print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
 226                     l3proto, l4proto);
 227
 228         if (seq_print_acct(s, ct, IP_CT_DIR_REPLY))
 229                 goto release;
 230
 231         if (test_bit(IPS_ASSURED_BIT, &ct->status))
 232                 seq_printf(s, "[ASSURED] ");
 233
 234         if (seq_has_overflowed(s))
 235                 goto release;
 236
 237 #if defined(CONFIG_NF_CONNTRACK_MARK)
 238         seq_printf(s, "mark=%u ", ct->mark);
 239 #endif
 240
 241         ct_show_secctx(s, ct);
 242         ct_show_zone(s, ct);
 243         ct_show_delta_time(s, ct);
 244
 245         seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use));
 246
 247         if (seq_has_overflowed(s))
 248                 goto release;
 249
 250         ret = 0;
 251 release:
 252         nf_ct_put(ct);
 253         return ret;
 254 }
 255
 256 static const struct seq_operations ct_seq_ops = {
 257         .start = ct_seq_start,
 258         .next  = ct_seq_next,
 259         .stop  = ct_seq_stop,
 260         .show  = ct_seq_show
 261 };
 262
 263 static int ct_open(struct inode *inode, struct file *file)
 264 {
 265         return seq_open_net(inode, file, &ct_seq_ops,
 266                         sizeof(struct ct_iter_state));
 267 }
 268
 269 static const struct file_operations ct_file_ops = {
 270         .owner   = THIS_MODULE,
 271         .open    = ct_open,
 272         .read    = seq_read,
 273         .llseek  = seq_lseek,
 274         .release = seq_release_net,
 275 };
 276
 277 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 278 {
 279         struct net *net = seq_file_net(seq);
 280         int cpu;
 281
 282         if (*pos == 0)
 283                 return SEQ_START_TOKEN;
 284
 285         for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
 286                 if (!cpu_possible(cpu))
 287                         continue;
 288                 *pos = cpu + 1;
 289                 return per_cpu_ptr(net->ct.stat, cpu);
 290         }
 291
 292         return NULL;
 293 }
 294
 295 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 296 {
 297         struct net *net = seq_file_net(seq);
 298         int cpu;
 299
 300         for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
 301                 if (!cpu_possible(cpu))
 302                         continue;
 303                 *pos = cpu + 1;
 304                 return per_cpu_ptr(net->ct.stat, cpu);
 305         }
 306
 307         return NULL;
 308 }
 309
 310 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 311 {
 312 }
 313
 314 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 315 {
 316         struct net *net = seq_file_net(seq);
 317         unsigned int nr_conntracks = atomic_read(&net->ct.count);
 318         const struct ip_conntrack_stat *st = v;
 319
 320         if (v == SEQ_START_TOKEN) {
 321                 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 search_restart\n");
 322                 return 0;
 323         }
 324
 325         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 326                         "%08x %08x %08x %08x %08x  %08x %08x %08x %08x\n",
 327                    nr_conntracks,
 328                    st->searched,
 329                    st->found,
 330                    st->new,
 331                    st->invalid,
 332                    st->ignore,
 333                    st->delete,
 334                    st->delete_list,
 335                    st->insert,
 336                    st->insert_failed,
 337                    st->drop,
 338                    st->early_drop,
 339                    st->error,
 340
 341                    st->expect_new,
 342                    st->expect_create,
 343                    st->expect_delete,
 344                    st->search_restart
 345                 );
 346         return 0;
 347 }
 348
 349 static const struct seq_operations ct_cpu_seq_ops = {
 350         .start  = ct_cpu_seq_start,
 351         .next   = ct_cpu_seq_next,
 352         .stop   = ct_cpu_seq_stop,
 353         .show   = ct_cpu_seq_show,
 354 };
 355
 356 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 357 {
 358         return seq_open_net(inode, file, &ct_cpu_seq_ops,
 359                             sizeof(struct seq_net_private));
 360 }
 361
 362 static const 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_net,
 368 };
 369
 370 static int nf_conntrack_standalone_init_proc(struct net *net)
 371 {
 372         struct proc_dir_entry *pde;
 373
 374         pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops);
 375         if (!pde)
 376                 goto out_nf_conntrack;
 377
 378         pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat,
 379                           &ct_cpu_seq_fops);
 380         if (!pde)
 381                 goto out_stat_nf_conntrack;
 382         return 0;
 383
 384 out_stat_nf_conntrack:
 385         remove_proc_entry("nf_conntrack", net->proc_net);
 386 out_nf_conntrack:
 387         return -ENOMEM;
 388 }
 389
 390 static void nf_conntrack_standalone_fini_proc(struct net *net)
 391 {
 392         remove_proc_entry("nf_conntrack", net->proc_net_stat);
 393         remove_proc_entry("nf_conntrack", net->proc_net);
 394 }
 395 #else
 396 static int nf_conntrack_standalone_init_proc(struct net *net)
 397 {
 398         return 0;
 399 }
 400
 401 static void nf_conntrack_standalone_fini_proc(struct net *net)
 402 {
 403 }
 404 #endif /* CONFIG_NF_CONNTRACK_PROCFS */
 405
 406 /* Sysctl support */
 407
 408 #ifdef CONFIG_SYSCTL
 409 /* Log invalid packets of a given protocol */
 410 static int log_invalid_proto_min = 0;
 411 static int log_invalid_proto_max = 255;
 412
 413 static struct ctl_table_header *nf_ct_netfilter_header;
 414
 415 static struct ctl_table nf_ct_sysctl_table[] = {
 416         {
 417                 .procname       = "nf_conntrack_max",
 418                 .data           = &nf_conntrack_max,
 419                 .maxlen         = sizeof(int),
 420                 .mode           = 0644,
 421                 .proc_handler   = proc_dointvec,
 422         },
 423         {
 424                 .procname       = "nf_conntrack_count",
 425                 .data           = &init_net.ct.count,
 426                 .maxlen         = sizeof(int),
 427                 .mode           = 0444,
 428                 .proc_handler   = proc_dointvec,
 429         },
 430         {
 431                 .procname       = "nf_conntrack_buckets",
 432                 .data           = &init_net.ct.htable_size,
 433                 .maxlen         = sizeof(unsigned int),
 434                 .mode           = 0444,
 435                 .proc_handler   = proc_dointvec,
 436         },
 437         {
 438                 .procname       = "nf_conntrack_checksum",
 439                 .data           = &init_net.ct.sysctl_checksum,
 440                 .maxlen         = sizeof(unsigned int),
 441                 .mode           = 0644,
 442                 .proc_handler   = proc_dointvec,
 443         },
 444         {
 445                 .procname       = "nf_conntrack_log_invalid",
 446                 .data           = &init_net.ct.sysctl_log_invalid,
 447                 .maxlen         = sizeof(unsigned int),
 448                 .mode           = 0644,
 449                 .proc_handler   = proc_dointvec_minmax,
 450                 .extra1         = &log_invalid_proto_min,
 451                 .extra2         = &log_invalid_proto_max,
 452         },
 453         {
 454                 .procname       = "nf_conntrack_expect_max",
 455                 .data           = &nf_ct_expect_max,
 456                 .maxlen         = sizeof(int),
 457                 .mode           = 0644,
 458                 .proc_handler   = proc_dointvec,
 459         },
 460         { }
 461 };
 462
 463 #define NET_NF_CONNTRACK_MAX 2089
 464
 465 static struct ctl_table nf_ct_netfilter_table[] = {
 466         {
 467                 .procname       = "nf_conntrack_max",
 468                 .data           = &nf_conntrack_max,
 469                 .maxlen         = sizeof(int),
 470                 .mode           = 0644,
 471                 .proc_handler   = proc_dointvec,
 472         },
 473         { }
 474 };
 475
 476 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 477 {
 478         struct ctl_table *table;
 479
 480         table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
 481                         GFP_KERNEL);
 482         if (!table)
 483                 goto out_kmemdup;
 484
 485         table[1].data = &net->ct.count;
 486         table[2].data = &net->ct.htable_size;
 487         table[3].data = &net->ct.sysctl_checksum;
 488         table[4].data = &net->ct.sysctl_log_invalid;
 489
 490         /* Don't export sysctls to unprivileged users */
 491         if (net->user_ns != &init_user_ns)
 492                 table[0].procname = NULL;
 493
 494         net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
 495         if (!net->ct.sysctl_header)
 496                 goto out_unregister_netfilter;
 497
 498         return 0;
 499
 500 out_unregister_netfilter:
 501         kfree(table);
 502 out_kmemdup:
 503         return -ENOMEM;
 504 }
 505
 506 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 507 {
 508         struct ctl_table *table;
 509
 510         table = net->ct.sysctl_header->ctl_table_arg;
 511         unregister_net_sysctl_table(net->ct.sysctl_header);
 512         kfree(table);
 513 }
 514 #else
 515 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 516 {
 517         return 0;
 518 }
 519
 520 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 521 {
 522 }
 523 #endif /* CONFIG_SYSCTL */
 524
 525 static int nf_conntrack_pernet_init(struct net *net)
 526 {
 527         int ret;
 528
 529         ret = nf_conntrack_init_net(net);
 530         if (ret < 0)
 531                 goto out_init;
 532
 533         ret = nf_conntrack_standalone_init_proc(net);
 534         if (ret < 0)
 535                 goto out_proc;
 536
 537         net->ct.sysctl_checksum = 1;
 538         net->ct.sysctl_log_invalid = 0;
 539         ret = nf_conntrack_standalone_init_sysctl(net);
 540         if (ret < 0)
 541                 goto out_sysctl;
 542
 543         return 0;
 544
 545 out_sysctl:
 546         nf_conntrack_standalone_fini_proc(net);
 547 out_proc:
 548         nf_conntrack_cleanup_net(net);
 549 out_init:
 550         return ret;
 551 }
 552
 553 static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
 554 {
 555         struct net *net;
 556
 557         list_for_each_entry(net, net_exit_list, exit_list) {
 558                 nf_conntrack_standalone_fini_sysctl(net);
 559                 nf_conntrack_standalone_fini_proc(net);
 560         }
 561         nf_conntrack_cleanup_net_list(net_exit_list);
 562 }
 563
 564 static struct pernet_operations nf_conntrack_net_ops = {
 565         .init           = nf_conntrack_pernet_init,
 566         .exit_batch     = nf_conntrack_pernet_exit,
 567 };
 568
 569 static int __init nf_conntrack_standalone_init(void)
 570 {
 571         int ret = nf_conntrack_init_start();
 572         if (ret < 0)
 573                 goto out_start;
 574
 575 #ifdef CONFIG_SYSCTL
 576         nf_ct_netfilter_header =
 577                 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
 578         if (!nf_ct_netfilter_header) {
 579                 pr_err("nf_conntrack: can't register to sysctl.\n");
 580                 ret = -ENOMEM;
 581                 goto out_sysctl;
 582         }
 583 #endif
 584
 585         ret = register_pernet_subsys(&nf_conntrack_net_ops);
 586         if (ret < 0)
 587                 goto out_pernet;
 588
 589         nf_conntrack_init_end();
 590         return 0;
 591
 592 out_pernet:
 593 #ifdef CONFIG_SYSCTL
 594         unregister_net_sysctl_table(nf_ct_netfilter_header);
 595 out_sysctl:
 596 #endif
 597         nf_conntrack_cleanup_end();
 598 out_start:
 599         return ret;
 600 }
 601
 602 static void __exit nf_conntrack_standalone_fini(void)
 603 {
 604         nf_conntrack_cleanup_start();
 605         unregister_pernet_subsys(&nf_conntrack_net_ops);
 606 #ifdef CONFIG_SYSCTL
 607         unregister_net_sysctl_table(nf_ct_netfilter_header);
 608 #endif
 609         nf_conntrack_cleanup_end();
 610 }
 611
 612 module_init(nf_conntrack_standalone_init);
 613 module_exit(nf_conntrack_standalone_fini);
 614
 615 /* Some modules need us, but don't depend directly on any symbol.
 616    They should call this. */
 617 void need_conntrack(void)
 618 {
 619 }
 620 EXPORT_SYMBOL_GPL(need_conntrack);