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                    nf_ct_expires(ct)  / HZ);
 232
 233         if (l4proto->print_conntrack)
 234                 l4proto->print_conntrack(s, ct);
 235
 236         print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 237                     l3proto, l4proto);
 238
 239         ct_show_zone(s, ct, NF_CT_ZONE_DIR_ORIG);
 240
 241         if (seq_has_overflowed(s))
 242                 goto release;
 243
 244         if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL))
 245                 goto release;
 246
 247         if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status)))
 248                 seq_printf(s, "[UNREPLIED] ");
 249
 250         print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
 251                     l3proto, l4proto);
 252
 253         ct_show_zone(s, ct, NF_CT_ZONE_DIR_REPL);
 254
 255         if (seq_print_acct(s, ct, IP_CT_DIR_REPLY))
 256                 goto release;
 257
 258         if (test_bit(IPS_ASSURED_BIT, &ct->status))
 259                 seq_printf(s, "[ASSURED] ");
 260
 261         if (seq_has_overflowed(s))
 262                 goto release;
 263
 264 #if defined(CONFIG_NF_CONNTRACK_MARK)
 265         seq_printf(s, "mark=%u ", ct->mark);
 266 #endif
 267
 268         ct_show_secctx(s, ct);
 269         ct_show_zone(s, ct, NF_CT_DEFAULT_ZONE_DIR);
 270         ct_show_delta_time(s, ct);
 271
 272         seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use));
 273
 274         if (seq_has_overflowed(s))
 275                 goto release;
 276
 277         ret = 0;
 278 release:
 279         nf_ct_put(ct);
 280         return ret;
 281 }
 282
 283 static const struct seq_operations ct_seq_ops = {
 284         .start = ct_seq_start,
 285         .next  = ct_seq_next,
 286         .stop  = ct_seq_stop,
 287         .show  = ct_seq_show
 288 };
 289
 290 static int ct_open(struct inode *inode, struct file *file)
 291 {
 292         return seq_open_net(inode, file, &ct_seq_ops,
 293                         sizeof(struct ct_iter_state));
 294 }
 295
 296 static const struct file_operations ct_file_ops = {
 297         .owner   = THIS_MODULE,
 298         .open    = ct_open,
 299         .read    = seq_read,
 300         .llseek  = seq_lseek,
 301         .release = seq_release_net,
 302 };
 303
 304 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 305 {
 306         struct net *net = seq_file_net(seq);
 307         int cpu;
 308
 309         if (*pos == 0)
 310                 return SEQ_START_TOKEN;
 311
 312         for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
 313                 if (!cpu_possible(cpu))
 314                         continue;
 315                 *pos = cpu + 1;
 316                 return per_cpu_ptr(net->ct.stat, cpu);
 317         }
 318
 319         return NULL;
 320 }
 321
 322 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 323 {
 324         struct net *net = seq_file_net(seq);
 325         int cpu;
 326
 327         for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
 328                 if (!cpu_possible(cpu))
 329                         continue;
 330                 *pos = cpu + 1;
 331                 return per_cpu_ptr(net->ct.stat, cpu);
 332         }
 333
 334         return NULL;
 335 }
 336
 337 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 338 {
 339 }
 340
 341 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 342 {
 343         struct net *net = seq_file_net(seq);
 344         unsigned int nr_conntracks = atomic_read(&net->ct.count);
 345         const struct ip_conntrack_stat *st = v;
 346
 347         if (v == SEQ_START_TOKEN) {
 348                 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");
 349                 return 0;
 350         }
 351
 352         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 353                         "%08x %08x %08x %08x %08x  %08x %08x %08x %08x\n",
 354                    nr_conntracks,
 355                    st->searched,
 356                    st->found,
 357                    st->new,
 358                    st->invalid,
 359                    st->ignore,
 360                    st->delete,
 361                    st->delete_list,
 362                    st->insert,
 363                    st->insert_failed,
 364                    st->drop,
 365                    st->early_drop,
 366                    st->error,
 367
 368                    st->expect_new,
 369                    st->expect_create,
 370                    st->expect_delete,
 371                    st->search_restart
 372                 );
 373         return 0;
 374 }
 375
 376 static const struct seq_operations ct_cpu_seq_ops = {
 377         .start  = ct_cpu_seq_start,
 378         .next   = ct_cpu_seq_next,
 379         .stop   = ct_cpu_seq_stop,
 380         .show   = ct_cpu_seq_show,
 381 };
 382
 383 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 384 {
 385         return seq_open_net(inode, file, &ct_cpu_seq_ops,
 386                             sizeof(struct seq_net_private));
 387 }
 388
 389 static const struct file_operations ct_cpu_seq_fops = {
 390         .owner   = THIS_MODULE,
 391         .open    = ct_cpu_seq_open,
 392         .read    = seq_read,
 393         .llseek  = seq_lseek,
 394         .release = seq_release_net,
 395 };
 396
 397 static int nf_conntrack_standalone_init_proc(struct net *net)
 398 {
 399         struct proc_dir_entry *pde;
 400         kuid_t root_uid;
 401         kgid_t root_gid;
 402
 403         pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops);
 404         if (!pde)
 405                 goto out_nf_conntrack;
 406
 407         root_uid = make_kuid(net->user_ns, 0);
 408         root_gid = make_kgid(net->user_ns, 0);
 409         if (uid_valid(root_uid) && gid_valid(root_gid))
 410                 proc_set_user(pde, root_uid, root_gid);
 411
 412         pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat,
 413                           &ct_cpu_seq_fops);
 414         if (!pde)
 415                 goto out_stat_nf_conntrack;
 416         return 0;
 417
 418 out_stat_nf_conntrack:
 419         remove_proc_entry("nf_conntrack", net->proc_net);
 420 out_nf_conntrack:
 421         return -ENOMEM;
 422 }
 423
 424 static void nf_conntrack_standalone_fini_proc(struct net *net)
 425 {
 426         remove_proc_entry("nf_conntrack", net->proc_net_stat);
 427         remove_proc_entry("nf_conntrack", net->proc_net);
 428 }
 429 #else
 430 static int nf_conntrack_standalone_init_proc(struct net *net)
 431 {
 432         return 0;
 433 }
 434
 435 static void nf_conntrack_standalone_fini_proc(struct net *net)
 436 {
 437 }
 438 #endif /* CONFIG_NF_CONNTRACK_PROCFS */
 439
 440 /* Sysctl support */
 441
 442 #ifdef CONFIG_SYSCTL
 443 /* Log invalid packets of a given protocol */
 444 static int log_invalid_proto_min __read_mostly;
 445 static int log_invalid_proto_max __read_mostly = 255;
 446
 447 /* size the user *wants to set */
 448 static unsigned int nf_conntrack_htable_size_user __read_mostly;
 449
 450 static int
 451 nf_conntrack_hash_sysctl(struct ctl_table *table, int write,
 452                          void __user *buffer, size_t *lenp, loff_t *ppos)
 453 {
 454         int ret;
 455
 456         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 457         if (ret < 0 || !write)
 458                 return ret;
 459
 460         /* update ret, we might not be able to satisfy request */
 461         ret = nf_conntrack_hash_resize(nf_conntrack_htable_size_user);
 462
 463         /* update it to the actual value used by conntrack */
 464         nf_conntrack_htable_size_user = nf_conntrack_htable_size;
 465         return ret;
 466 }
 467
 468 static struct ctl_table_header *nf_ct_netfilter_header;
 469
 470 static struct ctl_table nf_ct_sysctl_table[] = {
 471         {
 472                 .procname       = "nf_conntrack_max",
 473                 .data           = &nf_conntrack_max,
 474                 .maxlen         = sizeof(int),
 475                 .mode           = 0644,
 476                 .proc_handler   = proc_dointvec,
 477         },
 478         {
 479                 .procname       = "nf_conntrack_count",
 480                 .data           = &init_net.ct.count,
 481                 .maxlen         = sizeof(int),
 482                 .mode           = 0444,
 483                 .proc_handler   = proc_dointvec,
 484         },
 485         {
 486                 .procname       = "nf_conntrack_buckets",
 487                 .data           = &nf_conntrack_htable_size_user,
 488                 .maxlen         = sizeof(unsigned int),
 489                 .mode           = 0644,
 490                 .proc_handler   = nf_conntrack_hash_sysctl,
 491         },
 492         {
 493                 .procname       = "nf_conntrack_checksum",
 494                 .data           = &init_net.ct.sysctl_checksum,
 495                 .maxlen         = sizeof(unsigned int),
 496                 .mode           = 0644,
 497                 .proc_handler   = proc_dointvec,
 498         },
 499         {
 500                 .procname       = "nf_conntrack_log_invalid",
 501                 .data           = &init_net.ct.sysctl_log_invalid,
 502                 .maxlen         = sizeof(unsigned int),
 503                 .mode           = 0644,
 504                 .proc_handler   = proc_dointvec_minmax,
 505                 .extra1         = &log_invalid_proto_min,
 506                 .extra2         = &log_invalid_proto_max,
 507         },
 508         {
 509                 .procname       = "nf_conntrack_expect_max",
 510                 .data           = &nf_ct_expect_max,
 511                 .maxlen         = sizeof(int),
 512                 .mode           = 0644,
 513                 .proc_handler   = proc_dointvec,
 514         },
 515         { }
 516 };
 517
 518 static struct ctl_table nf_ct_netfilter_table[] = {
 519         {
 520                 .procname       = "nf_conntrack_max",
 521                 .data           = &nf_conntrack_max,
 522                 .maxlen         = sizeof(int),
 523                 .mode           = 0644,
 524                 .proc_handler   = proc_dointvec,
 525         },
 526         { }
 527 };
 528
 529 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 530 {
 531         struct ctl_table *table;
 532
 533         table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
 534                         GFP_KERNEL);
 535         if (!table)
 536                 goto out_kmemdup;
 537
 538         table[1].data = &net->ct.count;
 539         table[3].data = &net->ct.sysctl_checksum;
 540         table[4].data = &net->ct.sysctl_log_invalid;
 541
 542         /* Don't export sysctls to unprivileged users */
 543         if (net->user_ns != &init_user_ns)
 544                 table[0].procname = NULL;
 545
 546         if (!net_eq(&init_net, net))
 547                 table[2].mode = 0444;
 548
 549         net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
 550         if (!net->ct.sysctl_header)
 551                 goto out_unregister_netfilter;
 552
 553         return 0;
 554
 555 out_unregister_netfilter:
 556         kfree(table);
 557 out_kmemdup:
 558         return -ENOMEM;
 559 }
 560
 561 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 562 {
 563         struct ctl_table *table;
 564
 565         table = net->ct.sysctl_header->ctl_table_arg;
 566         unregister_net_sysctl_table(net->ct.sysctl_header);
 567         kfree(table);
 568 }
 569 #else
 570 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 571 {
 572         return 0;
 573 }
 574
 575 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 576 {
 577 }
 578 #endif /* CONFIG_SYSCTL */
 579
 580 static int nf_conntrack_pernet_init(struct net *net)
 581 {
 582         int ret;
 583
 584         ret = nf_conntrack_init_net(net);
 585         if (ret < 0)
 586                 goto out_init;
 587
 588         ret = nf_conntrack_standalone_init_proc(net);
 589         if (ret < 0)
 590                 goto out_proc;
 591
 592         net->ct.sysctl_checksum = 1;
 593         net->ct.sysctl_log_invalid = 0;
 594         ret = nf_conntrack_standalone_init_sysctl(net);
 595         if (ret < 0)
 596                 goto out_sysctl;
 597
 598         return 0;
 599
 600 out_sysctl:
 601         nf_conntrack_standalone_fini_proc(net);
 602 out_proc:
 603         nf_conntrack_cleanup_net(net);
 604 out_init:
 605         return ret;
 606 }
 607
 608 static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
 609 {
 610         struct net *net;
 611
 612         list_for_each_entry(net, net_exit_list, exit_list) {
 613                 nf_conntrack_standalone_fini_sysctl(net);
 614                 nf_conntrack_standalone_fini_proc(net);
 615         }
 616         nf_conntrack_cleanup_net_list(net_exit_list);
 617 }
 618
 619 static struct pernet_operations nf_conntrack_net_ops = {
 620         .init           = nf_conntrack_pernet_init,
 621         .exit_batch     = nf_conntrack_pernet_exit,
 622 };
 623
 624 static int __init nf_conntrack_standalone_init(void)
 625 {
 626         int ret = nf_conntrack_init_start();
 627         if (ret < 0)
 628                 goto out_start;
 629
 630 #ifdef CONFIG_SYSCTL
 631         nf_ct_netfilter_header =
 632                 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
 633         if (!nf_ct_netfilter_header) {
 634                 pr_err("nf_conntrack: can't register to sysctl.\n");
 635                 ret = -ENOMEM;
 636                 goto out_sysctl;
 637         }
 638
 639         nf_conntrack_htable_size_user = nf_conntrack_htable_size;
 640 #endif
 641
 642         ret = register_pernet_subsys(&nf_conntrack_net_ops);
 643         if (ret < 0)
 644                 goto out_pernet;
 645
 646         nf_conntrack_init_end();
 647         return 0;
 648
 649 out_pernet:
 650 #ifdef CONFIG_SYSCTL
 651         unregister_net_sysctl_table(nf_ct_netfilter_header);
 652 out_sysctl:
 653 #endif
 654         nf_conntrack_cleanup_end();
 655 out_start:
 656         return ret;
 657 }
 658
 659 static void __exit nf_conntrack_standalone_fini(void)
 660 {
 661         nf_conntrack_cleanup_start();
 662         unregister_pernet_subsys(&nf_conntrack_net_ops);
 663 #ifdef CONFIG_SYSCTL
 664         unregister_net_sysctl_table(nf_ct_netfilter_header);
 665 #endif
 666         nf_conntrack_cleanup_end();
 667 }
 668
 669 module_init(nf_conntrack_standalone_init);
 670 module_exit(nf_conntrack_standalone_fini);
 671
 672 /* Some modules need us, but don't depend directly on any symbol.
 673    They should call this. */
 674 void need_conntrack(void)
 675 {
 676 }
 677 EXPORT_SYMBOL_GPL(need_conntrack);