kernel/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                          int dir)
 146 {
 147         const struct nf_conntrack_zone *zone = nf_ct_zone(ct);
 148
 149         if (zone->dir != dir)
 150                 return;
 151         switch (zone->dir) {
 152         case NF_CT_DEFAULT_ZONE_DIR:
 153                 seq_printf(s, "zone=%u ", zone->id);
 154                 break;
 155         case NF_CT_ZONE_DIR_ORIG:
 156                 seq_printf(s, "zone-orig=%u ", zone->id);
 157                 break;
 158         case NF_CT_ZONE_DIR_REPL:
 159                 seq_printf(s, "zone-reply=%u ", zone->id);
 160                 break;
 161         default:
 162                 break;
 163         }
 164 }
 165 #else
 166 static inline void ct_show_zone(struct seq_file *s, const struct nf_conn *ct,
 167                                 int dir)
 168 {
 169 }
 170 #endif
 171
 172 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
 173 static void ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 174 {
 175         struct ct_iter_state *st = s->private;
 176         struct nf_conn_tstamp *tstamp;
 177         s64 delta_time;
 178
 179         tstamp = nf_conn_tstamp_find(ct);
 180         if (tstamp) {
 181                 delta_time = st->time_now - tstamp->start;
 182                 if (delta_time > 0)
 183                         delta_time = div_s64(delta_time, NSEC_PER_SEC);
 184                 else
 185                         delta_time = 0;
 186
 187                 seq_printf(s, "delta-time=%llu ",
 188                            (unsigned long long)delta_time);
 189         }
 190         return;
 191 }
 192 #else
 193 static inline void
 194 ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 195 {
 196 }
 197 #endif
 198
 199 /* return 0 on success, 1 in case of error */
 200 static int ct_seq_show(struct seq_file *s, void *v)
 201 {
 202         struct nf_conntrack_tuple_hash *hash = v;
 203         struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash);
 204         const struct nf_conntrack_l3proto *l3proto;
 205         const struct nf_conntrack_l4proto *l4proto;
 206         int ret = 0;
 207
 208         NF_CT_ASSERT(ct);
 209         if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
 210                 return 0;
 211
 212         /* we only want to print DIR_ORIGINAL */
 213         if (NF_CT_DIRECTION(hash))
 214                 goto release;
 215
 216         l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
 217         NF_CT_ASSERT(l3proto);
 218         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
 219         NF_CT_ASSERT(l4proto);
 220
 221         ret = -ENOSPC;
 222         seq_printf(s, "%-8s %u %-8s %u %ld ",
 223                    l3proto->name, nf_ct_l3num(ct),
 224                    l4proto->name, nf_ct_protonum(ct),
 225                    timer_pending(&ct->timeout)
 226                    ? (long)(ct->timeout.expires - jiffies)/HZ : 0);
 227
 228         if (l4proto->print_conntrack)
 229                 l4proto->print_conntrack(s, ct);
 230
 231         print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 232                     l3proto, l4proto);
 233
 234         ct_show_zone(s, ct, NF_CT_ZONE_DIR_ORIG);
 235
 236         if (seq_has_overflowed(s))
 237                 goto release;
 238
 239         if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL))
 240                 goto release;
 241
 242         if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status)))
 243                 seq_printf(s, "[UNREPLIED] ");
 244
 245         print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
 246                     l3proto, l4proto);
 247
 248         ct_show_zone(s, ct, NF_CT_ZONE_DIR_REPL);
 249
 250         if (seq_print_acct(s, ct, IP_CT_DIR_REPLY))
 251                 goto release;
 252
 253         if (test_bit(IPS_ASSURED_BIT, &ct->status))
 254                 seq_printf(s, "[ASSURED] ");
 255
 256         if (seq_has_overflowed(s))
 257                 goto release;
 258
 259 #if defined(CONFIG_NF_CONNTRACK_MARK)
 260         seq_printf(s, "mark=%u ", ct->mark);
 261 #endif
 262
 263         ct_show_secctx(s, ct);
 264         ct_show_zone(s, ct, NF_CT_DEFAULT_ZONE_DIR);
 265         ct_show_delta_time(s, ct);
 266
 267         seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use));
 268
 269         if (seq_has_overflowed(s))
 270                 goto release;
 271
 272         ret = 0;
 273 release:
 274         nf_ct_put(ct);
 275         return ret;
 276 }
 277
 278 static const struct seq_operations ct_seq_ops = {
 279         .start = ct_seq_start,
 280         .next  = ct_seq_next,
 281         .stop  = ct_seq_stop,
 282         .show  = ct_seq_show
 283 };
 284
 285 static int ct_open(struct inode *inode, struct file *file)
 286 {
 287         return seq_open_net(inode, file, &ct_seq_ops,
 288                         sizeof(struct ct_iter_state));
 289 }
 290
 291 static const struct file_operations ct_file_ops = {
 292         .owner   = THIS_MODULE,
 293         .open    = ct_open,
 294         .read    = seq_read,
 295         .llseek  = seq_lseek,
 296         .release = seq_release_net,
 297 };
 298
 299 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 300 {
 301         struct net *net = seq_file_net(seq);
 302         int cpu;
 303
 304         if (*pos == 0)
 305                 return SEQ_START_TOKEN;
 306
 307         for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
 308                 if (!cpu_possible(cpu))
 309                         continue;
 310                 *pos = cpu + 1;
 311                 return per_cpu_ptr(net->ct.stat, cpu);
 312         }
 313
 314         return NULL;
 315 }
 316
 317 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 318 {
 319         struct net *net = seq_file_net(seq);
 320         int cpu;
 321
 322         for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
 323                 if (!cpu_possible(cpu))
 324                         continue;
 325                 *pos = cpu + 1;
 326                 return per_cpu_ptr(net->ct.stat, cpu);
 327         }
 328
 329         return NULL;
 330 }
 331
 332 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 333 {
 334 }
 335
 336 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 337 {
 338         struct net *net = seq_file_net(seq);
 339         unsigned int nr_conntracks = atomic_read(&net->ct.count);
 340         const struct ip_conntrack_stat *st = v;
 341
 342         if (v == SEQ_START_TOKEN) {
 343                 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");
 344                 return 0;
 345         }
 346
 347         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 348                         "%08x %08x %08x %08x %08x  %08x %08x %08x %08x\n",
 349                    nr_conntracks,
 350                    st->searched,
 351                    st->found,
 352                    st->new,
 353                    st->invalid,
 354                    st->ignore,
 355                    st->delete,
 356                    st->delete_list,
 357                    st->insert,
 358                    st->insert_failed,
 359                    st->drop,
 360                    st->early_drop,
 361                    st->error,
 362
 363                    st->expect_new,
 364                    st->expect_create,
 365                    st->expect_delete,
 366                    st->search_restart
 367                 );
 368         return 0;
 369 }
 370
 371 static const struct seq_operations ct_cpu_seq_ops = {
 372         .start  = ct_cpu_seq_start,
 373         .next   = ct_cpu_seq_next,
 374         .stop   = ct_cpu_seq_stop,
 375         .show   = ct_cpu_seq_show,
 376 };
 377
 378 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 379 {
 380         return seq_open_net(inode, file, &ct_cpu_seq_ops,
 381                             sizeof(struct seq_net_private));
 382 }
 383
 384 static const struct file_operations ct_cpu_seq_fops = {
 385         .owner   = THIS_MODULE,
 386         .open    = ct_cpu_seq_open,
 387         .read    = seq_read,
 388         .llseek  = seq_lseek,
 389         .release = seq_release_net,
 390 };
 391
 392 static int nf_conntrack_standalone_init_proc(struct net *net)
 393 {
 394         struct proc_dir_entry *pde;
 395
 396         pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops);
 397         if (!pde)
 398                 goto out_nf_conntrack;
 399
 400         pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat,
 401                           &ct_cpu_seq_fops);
 402         if (!pde)
 403                 goto out_stat_nf_conntrack;
 404         return 0;
 405
 406 out_stat_nf_conntrack:
 407         remove_proc_entry("nf_conntrack", net->proc_net);
 408 out_nf_conntrack:
 409         return -ENOMEM;
 410 }
 411
 412 static void nf_conntrack_standalone_fini_proc(struct net *net)
 413 {
 414         remove_proc_entry("nf_conntrack", net->proc_net_stat);
 415         remove_proc_entry("nf_conntrack", net->proc_net);
 416 }
 417 #else
 418 static int nf_conntrack_standalone_init_proc(struct net *net)
 419 {
 420         return 0;
 421 }
 422
 423 static void nf_conntrack_standalone_fini_proc(struct net *net)
 424 {
 425 }
 426 #endif /* CONFIG_NF_CONNTRACK_PROCFS */
 427
 428 /* Sysctl support */
 429
 430 #ifdef CONFIG_SYSCTL
 431 /* Log invalid packets of a given protocol */
 432 static int log_invalid_proto_min = 0;
 433 static int log_invalid_proto_max = 255;
 434
 435 static struct ctl_table_header *nf_ct_netfilter_header;
 436
 437 static struct ctl_table nf_ct_sysctl_table[] = {
 438         {
 439                 .procname       = "nf_conntrack_max",
 440                 .data           = &nf_conntrack_max,
 441                 .maxlen         = sizeof(int),
 442                 .mode           = 0644,
 443                 .proc_handler   = proc_dointvec,
 444         },
 445         {
 446                 .procname       = "nf_conntrack_count",
 447                 .data           = &init_net.ct.count,
 448                 .maxlen         = sizeof(int),
 449                 .mode           = 0444,
 450                 .proc_handler   = proc_dointvec,
 451         },
 452         {
 453                 .procname       = "nf_conntrack_buckets",
 454                 .data           = &init_net.ct.htable_size,
 455                 .maxlen         = sizeof(unsigned int),
 456                 .mode           = 0444,
 457                 .proc_handler   = proc_dointvec,
 458         },
 459         {
 460                 .procname       = "nf_conntrack_checksum",
 461                 .data           = &init_net.ct.sysctl_checksum,
 462                 .maxlen         = sizeof(unsigned int),
 463                 .mode           = 0644,
 464                 .proc_handler   = proc_dointvec,
 465         },
 466         {
 467                 .procname       = "nf_conntrack_log_invalid",
 468                 .data           = &init_net.ct.sysctl_log_invalid,
 469                 .maxlen         = sizeof(unsigned int),
 470                 .mode           = 0644,
 471                 .proc_handler   = proc_dointvec_minmax,
 472                 .extra1         = &log_invalid_proto_min,
 473                 .extra2         = &log_invalid_proto_max,
 474         },
 475         {
 476                 .procname       = "nf_conntrack_expect_max",
 477                 .data           = &nf_ct_expect_max,
 478                 .maxlen         = sizeof(int),
 479                 .mode           = 0644,
 480                 .proc_handler   = proc_dointvec,
 481         },
 482         { }
 483 };
 484
 485 #define NET_NF_CONNTRACK_MAX 2089
 486
 487 static struct ctl_table nf_ct_netfilter_table[] = {
 488         {
 489                 .procname       = "nf_conntrack_max",
 490                 .data           = &nf_conntrack_max,
 491                 .maxlen         = sizeof(int),
 492                 .mode           = 0644,
 493                 .proc_handler   = proc_dointvec,
 494         },
 495         { }
 496 };
 497
 498 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 499 {
 500         struct ctl_table *table;
 501
 502         table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
 503                         GFP_KERNEL);
 504         if (!table)
 505                 goto out_kmemdup;
 506
 507         table[1].data = &net->ct.count;
 508         table[2].data = &net->ct.htable_size;
 509         table[3].data = &net->ct.sysctl_checksum;
 510         table[4].data = &net->ct.sysctl_log_invalid;
 511
 512         /* Don't export sysctls to unprivileged users */
 513         if (net->user_ns != &init_user_ns)
 514                 table[0].procname = NULL;
 515
 516         net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
 517         if (!net->ct.sysctl_header)
 518                 goto out_unregister_netfilter;
 519
 520         return 0;
 521
 522 out_unregister_netfilter:
 523         kfree(table);
 524 out_kmemdup:
 525         return -ENOMEM;
 526 }
 527
 528 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 529 {
 530         struct ctl_table *table;
 531
 532         table = net->ct.sysctl_header->ctl_table_arg;
 533         unregister_net_sysctl_table(net->ct.sysctl_header);
 534         kfree(table);
 535 }
 536 #else
 537 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 538 {
 539         return 0;
 540 }
 541
 542 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 543 {
 544 }
 545 #endif /* CONFIG_SYSCTL */
 546
 547 static int nf_conntrack_pernet_init(struct net *net)
 548 {
 549         int ret;
 550
 551         ret = nf_conntrack_init_net(net);
 552         if (ret < 0)
 553                 goto out_init;
 554
 555         ret = nf_conntrack_standalone_init_proc(net);
 556         if (ret < 0)
 557                 goto out_proc;
 558
 559         net->ct.sysctl_checksum = 1;
 560         net->ct.sysctl_log_invalid = 0;
 561         ret = nf_conntrack_standalone_init_sysctl(net);
 562         if (ret < 0)
 563                 goto out_sysctl;
 564
 565         return 0;
 566
 567 out_sysctl:
 568         nf_conntrack_standalone_fini_proc(net);
 569 out_proc:
 570         nf_conntrack_cleanup_net(net);
 571 out_init:
 572         return ret;
 573 }
 574
 575 static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
 576 {
 577         struct net *net;
 578
 579         list_for_each_entry(net, net_exit_list, exit_list) {
 580                 nf_conntrack_standalone_fini_sysctl(net);
 581                 nf_conntrack_standalone_fini_proc(net);
 582         }
 583         nf_conntrack_cleanup_net_list(net_exit_list);
 584 }
 585
 586 static struct pernet_operations nf_conntrack_net_ops = {
 587         .init           = nf_conntrack_pernet_init,
 588         .exit_batch     = nf_conntrack_pernet_exit,
 589 };
 590
 591 static int __init nf_conntrack_standalone_init(void)
 592 {
 593         int ret = nf_conntrack_init_start();
 594         if (ret < 0)
 595                 goto out_start;
 596
 597 #ifdef CONFIG_SYSCTL
 598         nf_ct_netfilter_header =
 599                 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
 600         if (!nf_ct_netfilter_header) {
 601                 pr_err("nf_conntrack: can't register to sysctl.\n");
 602                 ret = -ENOMEM;
 603                 goto out_sysctl;
 604         }
 605 #endif
 606
 607         ret = register_pernet_subsys(&nf_conntrack_net_ops);
 608         if (ret < 0)
 609                 goto out_pernet;
 610
 611         nf_conntrack_init_end();
 612         return 0;
 613
 614 out_pernet:
 615 #ifdef CONFIG_SYSCTL
 616         unregister_net_sysctl_table(nf_ct_netfilter_header);
 617 out_sysctl:
 618 #endif
 619         nf_conntrack_cleanup_end();
 620 out_start:
 621         return ret;
 622 }
 623
 624 static void __exit nf_conntrack_standalone_fini(void)
 625 {
 626         nf_conntrack_cleanup_start();
 627         unregister_pernet_subsys(&nf_conntrack_net_ops);
 628 #ifdef CONFIG_SYSCTL
 629         unregister_net_sysctl_table(nf_ct_netfilter_header);
 630 #endif
 631         nf_conntrack_cleanup_end();
 632 }
 633
 634 module_init(nf_conntrack_standalone_init);
 635 module_exit(nf_conntrack_standalone_fini);
 636
 637 /* Some modules need us, but don't depend directly on any symbol.
 638    They should call this. */
 639 void need_conntrack(void)
 640 {
 641 }
 642 EXPORT_SYMBOL_GPL(need_conntrack);