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 ct_iter_state *st = seq->private;
  58         struct hlist_nulls_node *n;
  59
  60         for (st->bucket = 0;
  61              st->bucket < nf_conntrack_htable_size;
  62              st->bucket++) {
  63                 n = rcu_dereference(hlist_nulls_first_rcu(&nf_conntrack_hash[st->bucket]));
  64                 if (!is_a_nulls(n))
  65                         return n;
  66         }
  67         return NULL;
  68 }
  69
  70 static struct hlist_nulls_node *ct_get_next(struct seq_file *seq,
  71                                       struct hlist_nulls_node *head)
  72 {
  73         struct ct_iter_state *st = seq->private;
  74
  75         head = rcu_dereference(hlist_nulls_next_rcu(head));
  76         while (is_a_nulls(head)) {
  77                 if (likely(get_nulls_value(head) == st->bucket)) {
  78                         if (++st->bucket >= nf_conntrack_htable_size)
  79                                 return NULL;
  80                 }
  81                 head = rcu_dereference(
  82                                 hlist_nulls_first_rcu(
  83                                         &nf_conntrack_hash[st->bucket]));
  84         }
  85         return head;
  86 }
  87
  88 static struct hlist_nulls_node *ct_get_idx(struct seq_file *seq, loff_t pos)
  89 {
  90         struct hlist_nulls_node *head = ct_get_first(seq);
  91
  92         if (head)
  93                 while (pos && (head = ct_get_next(seq, head)))
  94                         pos--;
  95         return pos ? NULL : head;
  96 }
  97
  98 static void *ct_seq_start(struct seq_file *seq, loff_t *pos)
  99         __acquires(RCU)
 100 {
 101         struct ct_iter_state *st = seq->private;
 102
 103         st->time_now = ktime_get_real_ns();
 104         rcu_read_lock();
 105         return ct_get_idx(seq, *pos);
 106 }
 107
 108 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
 109 {
 110         (*pos)++;
 111         return ct_get_next(s, v);
 112 }
 113
 114 static void ct_seq_stop(struct seq_file *s, void *v)
 115         __releases(RCU)
 116 {
 117         rcu_read_unlock();
 118 }
 119
 120 #ifdef CONFIG_NF_CONNTRACK_SECMARK
 121 static void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
 122 {
 123         int ret;
 124         u32 len;
 125         char *secctx;
 126
 127         ret = security_secid_to_secctx(ct->secmark, &secctx, &len);
 128         if (ret)
 129                 return;
 130
 131         seq_printf(s, "secctx=%s ", secctx);
 132
 133         security_release_secctx(secctx, len);
 134 }
 135 #else
 136 static inline void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
 137 {
 138 }
 139 #endif
 140
 141 #ifdef CONFIG_NF_CONNTRACK_ZONES
 142 static void ct_show_zone(struct seq_file *s, const struct nf_conn *ct,
 143                          int dir)
 144 {
 145         const struct nf_conntrack_zone *zone = nf_ct_zone(ct);
 146
 147         if (zone->dir != dir)
 148                 return;
 149         switch (zone->dir) {
 150         case NF_CT_DEFAULT_ZONE_DIR:
 151                 seq_printf(s, "zone=%u ", zone->id);
 152                 break;
 153         case NF_CT_ZONE_DIR_ORIG:
 154                 seq_printf(s, "zone-orig=%u ", zone->id);
 155                 break;
 156         case NF_CT_ZONE_DIR_REPL:
 157                 seq_printf(s, "zone-reply=%u ", zone->id);
 158                 break;
 159         default:
 160                 break;
 161         }
 162 }
 163 #else
 164 static inline void ct_show_zone(struct seq_file *s, const struct nf_conn *ct,
 165                                 int dir)
 166 {
 167 }
 168 #endif
 169
 170 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
 171 static void ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 172 {
 173         struct ct_iter_state *st = s->private;
 174         struct nf_conn_tstamp *tstamp;
 175         s64 delta_time;
 176
 177         tstamp = nf_conn_tstamp_find(ct);
 178         if (tstamp) {
 179                 delta_time = st->time_now - tstamp->start;
 180                 if (delta_time > 0)
 181                         delta_time = div_s64(delta_time, NSEC_PER_SEC);
 182                 else
 183                         delta_time = 0;
 184
 185                 seq_printf(s, "delta-time=%llu ",
 186                            (unsigned long long)delta_time);
 187         }
 188         return;
 189 }
 190 #else
 191 static inline void
 192 ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
 193 {
 194 }
 195 #endif
 196
 197 /* return 0 on success, 1 in case of error */
 198 static int ct_seq_show(struct seq_file *s, void *v)
 199 {
 200         struct nf_conntrack_tuple_hash *hash = v;
 201         struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash);
 202         const struct nf_conntrack_l3proto *l3proto;
 203         const struct nf_conntrack_l4proto *l4proto;
 204         int ret = 0;
 205
 206         NF_CT_ASSERT(ct);
 207         if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
 208                 return 0;
 209
 210         /* we only want to print DIR_ORIGINAL */
 211         if (NF_CT_DIRECTION(hash))
 212                 goto release;
 213
 214         l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
 215         NF_CT_ASSERT(l3proto);
 216         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
 217         NF_CT_ASSERT(l4proto);
 218
 219         ret = -ENOSPC;
 220         seq_printf(s, "%-8s %u %-8s %u %ld ",
 221                    l3proto->name, nf_ct_l3num(ct),
 222                    l4proto->name, nf_ct_protonum(ct),
 223                    timer_pending(&ct->timeout)
 224                    ? (long)(ct->timeout.expires - jiffies)/HZ : 0);
 225
 226         if (l4proto->print_conntrack)
 227                 l4proto->print_conntrack(s, ct);
 228
 229         print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 230                     l3proto, l4proto);
 231
 232         ct_show_zone(s, ct, NF_CT_ZONE_DIR_ORIG);
 233
 234         if (seq_has_overflowed(s))
 235                 goto release;
 236
 237         if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL))
 238                 goto release;
 239
 240         if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status)))
 241                 seq_printf(s, "[UNREPLIED] ");
 242
 243         print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
 244                     l3proto, l4proto);
 245
 246         ct_show_zone(s, ct, NF_CT_ZONE_DIR_REPL);
 247
 248         if (seq_print_acct(s, ct, IP_CT_DIR_REPLY))
 249                 goto release;
 250
 251         if (test_bit(IPS_ASSURED_BIT, &ct->status))
 252                 seq_printf(s, "[ASSURED] ");
 253
 254         if (seq_has_overflowed(s))
 255                 goto release;
 256
 257 #if defined(CONFIG_NF_CONNTRACK_MARK)
 258         seq_printf(s, "mark=%u ", ct->mark);
 259 #endif
 260
 261         ct_show_secctx(s, ct);
 262         ct_show_zone(s, ct, NF_CT_DEFAULT_ZONE_DIR);
 263         ct_show_delta_time(s, ct);
 264
 265         seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use));
 266
 267         if (seq_has_overflowed(s))
 268                 goto release;
 269
 270         ret = 0;
 271 release:
 272         nf_ct_put(ct);
 273         return ret;
 274 }
 275
 276 static const struct seq_operations ct_seq_ops = {
 277         .start = ct_seq_start,
 278         .next  = ct_seq_next,
 279         .stop  = ct_seq_stop,
 280         .show  = ct_seq_show
 281 };
 282
 283 static int ct_open(struct inode *inode, struct file *file)
 284 {
 285         return seq_open_net(inode, file, &ct_seq_ops,
 286                         sizeof(struct ct_iter_state));
 287 }
 288
 289 static const struct file_operations ct_file_ops = {
 290         .owner   = THIS_MODULE,
 291         .open    = ct_open,
 292         .read    = seq_read,
 293         .llseek  = seq_lseek,
 294         .release = seq_release_net,
 295 };
 296
 297 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
 298 {
 299         struct net *net = seq_file_net(seq);
 300         int cpu;
 301
 302         if (*pos == 0)
 303                 return SEQ_START_TOKEN;
 304
 305         for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
 306                 if (!cpu_possible(cpu))
 307                         continue;
 308                 *pos = cpu + 1;
 309                 return per_cpu_ptr(net->ct.stat, cpu);
 310         }
 311
 312         return NULL;
 313 }
 314
 315 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 316 {
 317         struct net *net = seq_file_net(seq);
 318         int cpu;
 319
 320         for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
 321                 if (!cpu_possible(cpu))
 322                         continue;
 323                 *pos = cpu + 1;
 324                 return per_cpu_ptr(net->ct.stat, cpu);
 325         }
 326
 327         return NULL;
 328 }
 329
 330 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
 331 {
 332 }
 333
 334 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
 335 {
 336         struct net *net = seq_file_net(seq);
 337         unsigned int nr_conntracks = atomic_read(&net->ct.count);
 338         const struct ip_conntrack_stat *st = v;
 339
 340         if (v == SEQ_START_TOKEN) {
 341                 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");
 342                 return 0;
 343         }
 344
 345         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
 346                         "%08x %08x %08x %08x %08x  %08x %08x %08x %08x\n",
 347                    nr_conntracks,
 348                    st->searched,
 349                    st->found,
 350                    st->new,
 351                    st->invalid,
 352                    st->ignore,
 353                    st->delete,
 354                    st->delete_list,
 355                    st->insert,
 356                    st->insert_failed,
 357                    st->drop,
 358                    st->early_drop,
 359                    st->error,
 360
 361                    st->expect_new,
 362                    st->expect_create,
 363                    st->expect_delete,
 364                    st->search_restart
 365                 );
 366         return 0;
 367 }
 368
 369 static const struct seq_operations ct_cpu_seq_ops = {
 370         .start  = ct_cpu_seq_start,
 371         .next   = ct_cpu_seq_next,
 372         .stop   = ct_cpu_seq_stop,
 373         .show   = ct_cpu_seq_show,
 374 };
 375
 376 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
 377 {
 378         return seq_open_net(inode, file, &ct_cpu_seq_ops,
 379                             sizeof(struct seq_net_private));
 380 }
 381
 382 static const struct file_operations ct_cpu_seq_fops = {
 383         .owner   = THIS_MODULE,
 384         .open    = ct_cpu_seq_open,
 385         .read    = seq_read,
 386         .llseek  = seq_lseek,
 387         .release = seq_release_net,
 388 };
 389
 390 static int nf_conntrack_standalone_init_proc(struct net *net)
 391 {
 392         struct proc_dir_entry *pde;
 393         kuid_t root_uid;
 394         kgid_t root_gid;
 395
 396         pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops);
 397         if (!pde)
 398                 goto out_nf_conntrack;
 399
 400         root_uid = make_kuid(net->user_ns, 0);
 401         root_gid = make_kgid(net->user_ns, 0);
 402         if (uid_valid(root_uid) && gid_valid(root_gid))
 403                 proc_set_user(pde, root_uid, root_gid);
 404
 405         pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat,
 406                           &ct_cpu_seq_fops);
 407         if (!pde)
 408                 goto out_stat_nf_conntrack;
 409         return 0;
 410
 411 out_stat_nf_conntrack:
 412         remove_proc_entry("nf_conntrack", net->proc_net);
 413 out_nf_conntrack:
 414         return -ENOMEM;
 415 }
 416
 417 static void nf_conntrack_standalone_fini_proc(struct net *net)
 418 {
 419         remove_proc_entry("nf_conntrack", net->proc_net_stat);
 420         remove_proc_entry("nf_conntrack", net->proc_net);
 421 }
 422 #else
 423 static int nf_conntrack_standalone_init_proc(struct net *net)
 424 {
 425         return 0;
 426 }
 427
 428 static void nf_conntrack_standalone_fini_proc(struct net *net)
 429 {
 430 }
 431 #endif /* CONFIG_NF_CONNTRACK_PROCFS */
 432
 433 /* Sysctl support */
 434
 435 #ifdef CONFIG_SYSCTL
 436 /* Log invalid packets of a given protocol */
 437 static int log_invalid_proto_min = 0;
 438 static int log_invalid_proto_max = 255;
 439
 440 static struct ctl_table_header *nf_ct_netfilter_header;
 441
 442 static struct ctl_table nf_ct_sysctl_table[] = {
 443         {
 444                 .procname       = "nf_conntrack_max",
 445                 .data           = &nf_conntrack_max,
 446                 .maxlen         = sizeof(int),
 447                 .mode           = 0644,
 448                 .proc_handler   = proc_dointvec,
 449         },
 450         {
 451                 .procname       = "nf_conntrack_count",
 452                 .data           = &init_net.ct.count,
 453                 .maxlen         = sizeof(int),
 454                 .mode           = 0444,
 455                 .proc_handler   = proc_dointvec,
 456         },
 457         {
 458                 .procname       = "nf_conntrack_buckets",
 459                 .data           = &nf_conntrack_htable_size,
 460                 .maxlen         = sizeof(unsigned int),
 461                 .mode           = 0444,
 462                 .proc_handler   = proc_dointvec,
 463         },
 464         {
 465                 .procname       = "nf_conntrack_checksum",
 466                 .data           = &init_net.ct.sysctl_checksum,
 467                 .maxlen         = sizeof(unsigned int),
 468                 .mode           = 0644,
 469                 .proc_handler   = proc_dointvec,
 470         },
 471         {
 472                 .procname       = "nf_conntrack_log_invalid",
 473                 .data           = &init_net.ct.sysctl_log_invalid,
 474                 .maxlen         = sizeof(unsigned int),
 475                 .mode           = 0644,
 476                 .proc_handler   = proc_dointvec_minmax,
 477                 .extra1         = &log_invalid_proto_min,
 478                 .extra2         = &log_invalid_proto_max,
 479         },
 480         {
 481                 .procname       = "nf_conntrack_expect_max",
 482                 .data           = &nf_ct_expect_max,
 483                 .maxlen         = sizeof(int),
 484                 .mode           = 0644,
 485                 .proc_handler   = proc_dointvec,
 486         },
 487         { }
 488 };
 489
 490 static struct ctl_table nf_ct_netfilter_table[] = {
 491         {
 492                 .procname       = "nf_conntrack_max",
 493                 .data           = &nf_conntrack_max,
 494                 .maxlen         = sizeof(int),
 495                 .mode           = 0644,
 496                 .proc_handler   = proc_dointvec,
 497         },
 498         { }
 499 };
 500
 501 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 502 {
 503         struct ctl_table *table;
 504
 505         table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
 506                         GFP_KERNEL);
 507         if (!table)
 508                 goto out_kmemdup;
 509
 510         table[1].data = &net->ct.count;
 511         table[3].data = &net->ct.sysctl_checksum;
 512         table[4].data = &net->ct.sysctl_log_invalid;
 513
 514         /* Don't export sysctls to unprivileged users */
 515         if (net->user_ns != &init_user_ns)
 516                 table[0].procname = NULL;
 517
 518         net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
 519         if (!net->ct.sysctl_header)
 520                 goto out_unregister_netfilter;
 521
 522         return 0;
 523
 524 out_unregister_netfilter:
 525         kfree(table);
 526 out_kmemdup:
 527         return -ENOMEM;
 528 }
 529
 530 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 531 {
 532         struct ctl_table *table;
 533
 534         table = net->ct.sysctl_header->ctl_table_arg;
 535         unregister_net_sysctl_table(net->ct.sysctl_header);
 536         kfree(table);
 537 }
 538 #else
 539 static int nf_conntrack_standalone_init_sysctl(struct net *net)
 540 {
 541         return 0;
 542 }
 543
 544 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
 545 {
 546 }
 547 #endif /* CONFIG_SYSCTL */
 548
 549 static int nf_conntrack_pernet_init(struct net *net)
 550 {
 551         int ret;
 552
 553         ret = nf_conntrack_init_net(net);
 554         if (ret < 0)
 555                 goto out_init;
 556
 557         ret = nf_conntrack_standalone_init_proc(net);
 558         if (ret < 0)
 559                 goto out_proc;
 560
 561         net->ct.sysctl_checksum = 1;
 562         net->ct.sysctl_log_invalid = 0;
 563         ret = nf_conntrack_standalone_init_sysctl(net);
 564         if (ret < 0)
 565                 goto out_sysctl;
 566
 567         return 0;
 568
 569 out_sysctl:
 570         nf_conntrack_standalone_fini_proc(net);
 571 out_proc:
 572         nf_conntrack_cleanup_net(net);
 573 out_init:
 574         return ret;
 575 }
 576
 577 static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
 578 {
 579         struct net *net;
 580
 581         list_for_each_entry(net, net_exit_list, exit_list) {
 582                 nf_conntrack_standalone_fini_sysctl(net);
 583                 nf_conntrack_standalone_fini_proc(net);
 584         }
 585         nf_conntrack_cleanup_net_list(net_exit_list);
 586 }
 587
 588 static struct pernet_operations nf_conntrack_net_ops = {
 589         .init           = nf_conntrack_pernet_init,
 590         .exit_batch     = nf_conntrack_pernet_exit,
 591 };
 592
 593 static int __init nf_conntrack_standalone_init(void)
 594 {
 595         int ret = nf_conntrack_init_start();
 596         if (ret < 0)
 597                 goto out_start;
 598
 599 #ifdef CONFIG_SYSCTL
 600         nf_ct_netfilter_header =
 601                 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
 602         if (!nf_ct_netfilter_header) {
 603                 pr_err("nf_conntrack: can't register to sysctl.\n");
 604                 ret = -ENOMEM;
 605                 goto out_sysctl;
 606         }
 607 #endif
 608
 609         ret = register_pernet_subsys(&nf_conntrack_net_ops);
 610         if (ret < 0)
 611                 goto out_pernet;
 612
 613         nf_conntrack_init_end();
 614         return 0;
 615
 616 out_pernet:
 617 #ifdef CONFIG_SYSCTL
 618         unregister_net_sysctl_table(nf_ct_netfilter_header);
 619 out_sysctl:
 620 #endif
 621         nf_conntrack_cleanup_end();
 622 out_start:
 623         return ret;
 624 }
 625
 626 static void __exit nf_conntrack_standalone_fini(void)
 627 {
 628         nf_conntrack_cleanup_start();
 629         unregister_pernet_subsys(&nf_conntrack_net_ops);
 630 #ifdef CONFIG_SYSCTL
 631         unregister_net_sysctl_table(nf_ct_netfilter_header);
 632 #endif
 633         nf_conntrack_cleanup_end();
 634 }
 635
 636 module_init(nf_conntrack_standalone_init);
 637 module_exit(nf_conntrack_standalone_fini);
 638
 639 /* Some modules need us, but don't depend directly on any symbol.
 640    They should call this. */
 641 void need_conntrack(void)
 642 {
 643 }
 644 EXPORT_SYMBOL_GPL(need_conntrack);