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