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         struct hlist_nulls_head *hash;
  52         unsigned int htable_size;
  53         unsigned int bucket;
  54         u_int64_t time_now;
  55 };
  56
  57 static struct hlist_nulls_node *ct_get_first(struct seq_file *seq)
  58 {
  59         struct ct_iter_state *st = seq->private;
  60         struct hlist_nulls_node *n;
  61
  62         for (st->bucket = 0;
  63              st->bucket < st->htable_size;
  64              st->bucket++) {
  65                 n = rcu_dereference(
  66                         hlist_nulls_first_rcu(&st->hash[st->bucket]));
  67                 if (!is_a_nulls(n))
  68                         return n;
  69         }
  70         return NULL;
  71 }
  72
  73 static struct hlist_nulls_node *ct_get_next(struct seq_file *seq,
  74                                       struct hlist_nulls_node *head)
  75 {
  76         struct ct_iter_state *st = seq->private;
  77
  78         head = rcu_dereference(hlist_nulls_next_rcu(head));
  79         while (is_a_nulls(head)) {
  80                 if (likely(get_nulls_value(head) == st->bucket)) {
  81                         if (++st->bucket >= st->htable_size)
  82                                 return NULL;
  83                 }
  84                 head = rcu_dereference(
  85                         hlist_nulls_first_rcu(&st->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
 108         nf_conntrack_get_ht(&st->hash, &st->htable_size);
 109         return ct_get_idx(seq, *pos);
 110 }
 111
 112 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
 113 {
 114         (*pos)++;
 115         return ct_get_next(s, v);
 116 }
 117
 118 static void ct_seq_stop(struct seq_file *s, void *v)
 119         __releases(RCU)
 120 {
 121         rcu_read_unlock();
 122 }
 123
 124 #ifdef CONFIG_NF_CONNTRACK_SECMARK
 125 static void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
 126 {
 127         int ret;
 128         u32 len;
 129         char *secctx;
 130
 131         ret = security_secid_to_secctx(ct->secmark, &secctx, &len);
 132         if (ret)
 133                 return;
 134
 135         seq_printf(s, "secctx=%s ", secctx);
 136
 137         security_release_secctx(secctx, len);
 138 }
 139 #else
 140 static inline void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
 141 {
 142 }
 143 #endif
 144
 145 #ifdef CONFIG_NF_CONNTRACK_ZONES
 146 static void ct_show_zone(struct seq_file *s, const struct nf_conn *ct,
 147                          int dir)
 148 {
 149         const struct nf_conntrack_zone *zone = nf_ct_zone(ct);
 150
 151         if (zone->dir != dir)
 152                 return;
 153         switch (zone->dir) {
 154         case NF_CT_DEFAULT_ZONE_DIR:
 155                 seq_printf(s, "zone=%u ", zone->id);
 156                 break;
 157         case NF_CT_ZONE_DIR_ORIG:
 158                 seq_printf(s, "zone-orig=%u ", zone->id);
 159                 break;
 160         case NF_CT_ZONE_DIR_REPL:
 161                 seq_printf(s, "zone-reply=%u ", zone->id);
 162                 break;
 163         default:
 164                 break;
 165         }
 166 }
 167 #else
 168 static inline void ct_show_zone(struct seq_file *s, const struct nf_conn *ct,
 169                                 int dir)
 170 {
 171 }
 172 #endif
 173
 174 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
 175 static void ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 176 {
 177         struct ct_iter_state *st = s->private;
 178         struct nf_conn_tstamp *tstamp;
 179         s64 delta_time;
 180
 181         tstamp = nf_conn_tstamp_find(ct);
 182         if (tstamp) {
 183                 delta_time = st->time_now - tstamp->start;
 184                 if (delta_time > 0)
 185                         delta_time = div_s64(delta_time, NSEC_PER_SEC);
 186                 else
 187                         delta_time = 0;
 188
 189                 seq_printf(s, "delta-time=%llu ",
 190                            (unsigned long long)delta_time);
 191         }
 192         return;
 193 }
 194 #else
 195 static inline void
 196 ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 197 {
 198 }
 199 #endif
 200
 201 /* return 0 on success, 1 in case of error */
 202 static int ct_seq_show(struct seq_file *s, void *v)
 203 {
 204         struct nf_conntrack_tuple_hash *hash = v;
 205         struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash);
 206         const struct nf_conntrack_l3proto *l3proto;
 207         const struct nf_conntrack_l4proto *l4proto;
 208         struct net *net = seq_file_net(s);
 209         int ret = 0;
 210
 211         NF_CT_ASSERT(ct);
 212         if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
 213                 return 0;
 214
 215         /* we only want to print DIR_ORIGINAL */
 216         if (NF_CT_DIRECTION(hash))
 217                 goto release;
 218
 219         if (!net_eq(nf_ct_net(ct), net))
 220                 goto release;
 221
 222         l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
 223         NF_CT_ASSERT(l3proto);
 224         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
 225         NF_CT_ASSERT(l4proto);
 226
 227         ret = -ENOSPC;
 228         seq_printf(s, "%-8s %u %-8s %u %ld ",
 229                    l3proto->name, nf_ct_l3num(ct),
 230                    l4proto->name, nf_ct_protonum(ct),
 231                    timer_pending(&ct->timeout)
 232                    ? (long)(ct->timeout.expires - jiffies)/HZ : 0);
 233
 234         if (l4proto->print_conntrack)
 235                 l4proto->print_conntrack(s, ct);
 236
 237         print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 238                     l3proto, l4proto);
 239
 240         ct_show_zone(s, ct, NF_CT_ZONE_DIR_ORIG);
 241
 242         if (seq_has_overflowed(s))
 243                 goto release;
 244
 245         if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL))
 246                 goto release;
 247
 248         if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status)))
 249                 seq_printf(s, "[UNREPLIED] ");
 250
 251         print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
 252                     l3proto, l4proto);
 253
 254         ct_show_zone(s, ct, NF_CT_ZONE_DIR_REPL);
 255
 256         if (seq_print_acct(s, ct, IP_CT_DIR_REPLY))
 257                 goto release;
 258
 259         if (test_bit(IPS_ASSURED_BIT, &ct->status))
 260                 seq_printf(s, "[ASSURED] ");
 261
 262         if (seq_has_overflowed(s))
 263                 goto release;
 264
 265 #if defined(CONFIG_NF_CONNTRACK_MARK)
 266         seq_printf(s, "mark=%u ", ct->mark);
 267 #endif
 268
 269         ct_show_secctx(s, ct);
 270         ct_show_zone(s, ct, NF_CT_DEFAULT_ZONE_DIR);
 271         ct_show_delta_time(s, ct);
 272
 273         seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use));
 274
 275         if (seq_has_overflowed(s))
 276                 goto release;
 277
 278         ret = 0;
 279 release:
 280         nf_ct_put(ct);
 281         return ret;
 282 }
 283
 284 static const struct seq_operations ct_seq_ops = {
 285         .start = ct_seq_start,
 286         .next  = ct_seq_next,
 287         .stop  = ct_seq_stop,
 288         .show  = ct_seq_show
 289 };
 290
 291 static int ct_open(struct inode *inode, struct file *file)
 292 {
 293         return seq_open_net(inode, file, &ct_seq_ops,
 294                         sizeof(struct ct_iter_state));
 295 }
 296
 297 static const struct file_operations ct_file_ops = {
 298         .owner   = THIS_MODULE,
 299         .open    = ct_open,
 300         .read    = seq_read,
 301         .llseek  = seq_lseek,
 302         .release = seq_release_net,
 303 };
 304
 305 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 306 {
 307         struct net *net = seq_file_net(seq);
 308         int cpu;
 309
 310         if (*pos == 0)
 311                 return SEQ_START_TOKEN;
 312
 313         for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
 314                 if (!cpu_possible(cpu))
 315                         continue;
 316                 *pos = cpu + 1;
 317                 return per_cpu_ptr(net->ct.stat, cpu);
 318         }
 319
 320         return NULL;
 321 }
 322
 323 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 324 {
 325         struct net *net = seq_file_net(seq);
 326         int cpu;
 327
 328         for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
 329                 if (!cpu_possible(cpu))
 330                         continue;
 331                 *pos = cpu + 1;
 332                 return per_cpu_ptr(net->ct.stat, cpu);
 333         }
 334
 335         return NULL;
 336 }
 337
 338 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 339 {
 340 }
 341
 342 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 343 {
 344         struct net *net = seq_file_net(seq);
 345         unsigned int nr_conntracks = atomic_read(&net->ct.count);
 346         const struct ip_conntrack_stat *st = v;
 347
 348         if (v == SEQ_START_TOKEN) {
 349                 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");
 350                 return 0;
 351         }
 352
 353         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 354                         "%08x %08x %08x %08x %08x  %08x %08x %08x %08x\n",
 355                    nr_conntracks,
 356                    st->searched,
 357                    st->found,
 358                    st->new,
 359                    st->invalid,
 360                    st->ignore,
 361                    st->delete,
 362                    st->delete_list,
 363                    st->insert,
 364                    st->insert_failed,
 365                    st->drop,
 366                    st->early_drop,
 367                    st->error,
 368
 369                    st->expect_new,
 370                    st->expect_create,
 371                    st->expect_delete,
 372                    st->search_restart
 373                 );
 374         return 0;
 375 }
 376
 377 static const struct seq_operations ct_cpu_seq_ops = {
 378         .start  = ct_cpu_seq_start,
 379         .next   = ct_cpu_seq_next,
 380         .stop   = ct_cpu_seq_stop,
 381         .show   = ct_cpu_seq_show,
 382 };
 383
 384 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 385 {
 386         return seq_open_net(inode, file, &ct_cpu_seq_ops,
 387                             sizeof(struct seq_net_private));
 388 }
 389
 390 static const struct file_operations ct_cpu_seq_fops = {
 391         .owner   = THIS_MODULE,
 392         .open    = ct_cpu_seq_open,
 393         .read    = seq_read,
 394         .llseek  = seq_lseek,
 395         .release = seq_release_net,
 396 };
 397
 398 static int nf_conntrack_standalone_init_proc(struct net *net)
 399 {
 400         struct proc_dir_entry *pde;
 401         kuid_t root_uid;
 402         kgid_t root_gid;
 403
 404         pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops);
 405         if (!pde)
 406                 goto out_nf_conntrack;
 407
 408         root_uid = make_kuid(net->user_ns, 0);
 409         root_gid = make_kgid(net->user_ns, 0);
 410         if (uid_valid(root_uid) && gid_valid(root_gid))
 411                 proc_set_user(pde, root_uid, root_gid);
 412
 413         pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat,
 414                           &ct_cpu_seq_fops);
 415         if (!pde)
 416                 goto out_stat_nf_conntrack;
 417         return 0;
 418
 419 out_stat_nf_conntrack:
 420         remove_proc_entry("nf_conntrack", net->proc_net);
 421 out_nf_conntrack:
 422         return -ENOMEM;
 423 }
 424
 425 static void nf_conntrack_standalone_fini_proc(struct net *net)
 426 {
 427         remove_proc_entry("nf_conntrack", net->proc_net_stat);
 428         remove_proc_entry("nf_conntrack", net->proc_net);
 429 }
 430 #else
 431 static int nf_conntrack_standalone_init_proc(struct net *net)
 432 {
 433         return 0;
 434 }
 435
 436 static void nf_conntrack_standalone_fini_proc(struct net *net)
 437 {
 438 }
 439 #endif /* CONFIG_NF_CONNTRACK_PROCFS */
 440
 441 /* Sysctl support */
 442
 443 #ifdef CONFIG_SYSCTL
 444 /* Log invalid packets of a given protocol */
 445 static int log_invalid_proto_min __read_mostly;
 446 static int log_invalid_proto_max __read_mostly = 255;
 447
 448 /* size the user *wants to set */
 449 static unsigned int nf_conntrack_htable_size_user __read_mostly;
 450
 451 static int
 452 nf_conntrack_hash_sysctl(struct ctl_table *table, int write,
 453                          void __user *buffer, size_t *lenp, loff_t *ppos)
 454 {
 455         int ret;
 456
 457         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 458         if (ret < 0 || !write)
 459                 return ret;
 460
 461         /* update ret, we might not be able to satisfy request */
 462         ret = nf_conntrack_hash_resize(nf_conntrack_htable_size_user);
 463
 464         /* update it to the actual value used by conntrack */
 465         nf_conntrack_htable_size_user = nf_conntrack_htable_size;
 466         return ret;
 467 }
 468
 469 static struct ctl_table_header *nf_ct_netfilter_header;
 470
 471 static struct ctl_table nf_ct_sysctl_table[] = {
 472         {
 473                 .procname       = "nf_conntrack_max",
 474                 .data           = &nf_conntrack_max,
 475                 .maxlen         = sizeof(int),
 476                 .mode           = 0644,
 477                 .proc_handler   = proc_dointvec,
 478         },
 479         {
 480                 .procname       = "nf_conntrack_count",
 481                 .data           = &init_net.ct.count,
 482                 .maxlen         = sizeof(int),
 483                 .mode           = 0444,
 484                 .proc_handler   = proc_dointvec,
 485         },
 486         {
 487                 .procname       = "nf_conntrack_buckets",
 488                 .data           = &nf_conntrack_htable_size_user,
 489                 .maxlen         = sizeof(unsigned int),
 490                 .mode           = 0644,
 491                 .proc_handler   = nf_conntrack_hash_sysctl,
 492         },
 493         {
 494                 .procname       = "nf_conntrack_checksum",
 495                 .data           = &init_net.ct.sysctl_checksum,
 496                 .maxlen         = sizeof(unsigned int),
 497                 .mode           = 0644,
 498                 .proc_handler   = proc_dointvec,
 499         },
 500         {
 501                 .procname       = "nf_conntrack_log_invalid",
 502                 .data           = &init_net.ct.sysctl_log_invalid,
 503                 .maxlen         = sizeof(unsigned int),
 504                 .mode           = 0644,
 505                 .proc_handler   = proc_dointvec_minmax,
 506                 .extra1         = &log_invalid_proto_min,
 507                 .extra2         = &log_invalid_proto_max,
 508         },
 509         {
 510                 .procname       = "nf_conntrack_expect_max",
 511                 .data           = &nf_ct_expect_max,
 512                 .maxlen         = sizeof(int),
 513                 .mode           = 0644,
 514                 .proc_handler   = proc_dointvec,
 515         },
 516         { }
 517 };
 518
 519 static struct ctl_table nf_ct_netfilter_table[] = {
 520         {
 521                 .procname       = "nf_conntrack_max",
 522                 .data           = &nf_conntrack_max,
 523                 .maxlen         = sizeof(int),
 524                 .mode           = 0644,
 525                 .proc_handler   = proc_dointvec,
 526         },
 527         { }
 528 };
 529
 530 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 531 {
 532         struct ctl_table *table;
 533
 534         table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
 535                         GFP_KERNEL);
 536         if (!table)
 537                 goto out_kmemdup;
 538
 539         table[1].data = &net->ct.count;
 540         table[3].data = &net->ct.sysctl_checksum;
 541         table[4].data = &net->ct.sysctl_log_invalid;
 542
 543         /* Don't export sysctls to unprivileged users */
 544         if (net->user_ns != &init_user_ns)
 545                 table[0].procname = NULL;
 546
 547         if (!net_eq(&init_net, net))
 548                 table[2].mode = 0444;
 549
 550         net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
 551         if (!net->ct.sysctl_header)
 552                 goto out_unregister_netfilter;
 553
 554         return 0;
 555
 556 out_unregister_netfilter:
 557         kfree(table);
 558 out_kmemdup:
 559         return -ENOMEM;
 560 }
 561
 562 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 563 {
 564         struct ctl_table *table;
 565
 566         table = net->ct.sysctl_header->ctl_table_arg;
 567         unregister_net_sysctl_table(net->ct.sysctl_header);
 568         kfree(table);
 569 }
 570 #else
 571 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 572 {
 573         return 0;
 574 }
 575
 576 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 577 {
 578 }
 579 #endif /* CONFIG_SYSCTL */
 580
 581 static int nf_conntrack_pernet_init(struct net *net)
 582 {
 583         int ret;
 584
 585         ret = nf_conntrack_init_net(net);
 586         if (ret < 0)
 587                 goto out_init;
 588
 589         ret = nf_conntrack_standalone_init_proc(net);
 590         if (ret < 0)
 591                 goto out_proc;
 592
 593         net->ct.sysctl_checksum = 1;
 594         net->ct.sysctl_log_invalid = 0;
 595         ret = nf_conntrack_standalone_init_sysctl(net);
 596         if (ret < 0)
 597                 goto out_sysctl;
 598
 599         return 0;
 600
 601 out_sysctl:
 602         nf_conntrack_standalone_fini_proc(net);
 603 out_proc:
 604         nf_conntrack_cleanup_net(net);
 605 out_init:
 606         return ret;
 607 }
 608
 609 static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
 610 {
 611         struct net *net;
 612
 613         list_for_each_entry(net, net_exit_list, exit_list) {
 614                 nf_conntrack_standalone_fini_sysctl(net);
 615                 nf_conntrack_standalone_fini_proc(net);
 616         }
 617         nf_conntrack_cleanup_net_list(net_exit_list);
 618 }
 619
 620 static struct pernet_operations nf_conntrack_net_ops = {
 621         .init           = nf_conntrack_pernet_init,
 622         .exit_batch     = nf_conntrack_pernet_exit,
 623 };
 624
 625 static int __init nf_conntrack_standalone_init(void)
 626 {
 627         int ret = nf_conntrack_init_start();
 628         if (ret < 0)
 629                 goto out_start;
 630
 631 #ifdef CONFIG_SYSCTL
 632         nf_ct_netfilter_header =
 633                 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
 634         if (!nf_ct_netfilter_header) {
 635                 pr_err("nf_conntrack: can't register to sysctl.\n");
 636                 ret = -ENOMEM;
 637                 goto out_sysctl;
 638         }
 639
 640         nf_conntrack_htable_size_user = nf_conntrack_htable_size;
 641 #endif
 642
 643         ret = register_pernet_subsys(&nf_conntrack_net_ops);
 644         if (ret < 0)
 645                 goto out_pernet;
 646
 647         nf_conntrack_init_end();
 648         return 0;
 649
 650 out_pernet:
 651 #ifdef CONFIG_SYSCTL
 652         unregister_net_sysctl_table(nf_ct_netfilter_header);
 653 out_sysctl:
 654 #endif
 655         nf_conntrack_cleanup_end();
 656 out_start:
 657         return ret;
 658 }
 659
 660 static void __exit nf_conntrack_standalone_fini(void)
 661 {
 662         nf_conntrack_cleanup_start();
 663         unregister_pernet_subsys(&nf_conntrack_net_ops);
 664 #ifdef CONFIG_SYSCTL
 665         unregister_net_sysctl_table(nf_ct_netfilter_header);
 666 #endif
 667         nf_conntrack_cleanup_end();
 668 }
 669
 670 module_init(nf_conntrack_standalone_init);
 671 module_exit(nf_conntrack_standalone_fini);
 672
 673 /* Some modules need us, but don't depend directly on any symbol.
 674    They should call this. */
 675 void need_conntrack(void)
 676 {
 677 }
 678 EXPORT_SYMBOL_GPL(need_conntrack);