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 if (nf_ct_should_gc(ct)) { 216 nf_ct_kill(ct); 217 goto release; 218 } 219 220 /* we only want to print DIR_ORIGINAL */ 221 if (NF_CT_DIRECTION(hash)) 222 goto release; 223 224 if (!net_eq(nf_ct_net(ct), net)) 225 goto release; 226 227 l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct)); 228 NF_CT_ASSERT(l3proto); 229 l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); 230 NF_CT_ASSERT(l4proto); 231 232 ret = -ENOSPC; 233 seq_printf(s, "%-8s %u %-8s %u %ld ", 234 l3proto->name, nf_ct_l3num(ct), 235 l4proto->name, nf_ct_protonum(ct), 236 nf_ct_expires(ct) / HZ); 237 238 if (l4proto->print_conntrack) 239 l4proto->print_conntrack(s, ct); 240 241 print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, 242 l3proto, l4proto); 243 244 ct_show_zone(s, ct, NF_CT_ZONE_DIR_ORIG); 245 246 if (seq_has_overflowed(s)) 247 goto release; 248 249 if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL)) 250 goto release; 251 252 if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status))) 253 seq_puts(s, "[UNREPLIED] "); 254 255 print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, 256 l3proto, l4proto); 257 258 ct_show_zone(s, ct, NF_CT_ZONE_DIR_REPL); 259 260 if (seq_print_acct(s, ct, IP_CT_DIR_REPLY)) 261 goto release; 262 263 if (test_bit(IPS_ASSURED_BIT, &ct->status)) 264 seq_puts(s, "[ASSURED] "); 265 266 if (seq_has_overflowed(s)) 267 goto release; 268 269 #if defined(CONFIG_NF_CONNTRACK_MARK) 270 seq_printf(s, "mark=%u ", ct->mark); 271 #endif 272 273 ct_show_secctx(s, ct); 274 ct_show_zone(s, ct, NF_CT_DEFAULT_ZONE_DIR); 275 ct_show_delta_time(s, ct); 276 277 seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use)); 278 279 if (seq_has_overflowed(s)) 280 goto release; 281 282 ret = 0; 283 release: 284 nf_ct_put(ct); 285 return ret; 286 } 287 288 static const struct seq_operations ct_seq_ops = { 289 .start = ct_seq_start, 290 .next = ct_seq_next, 291 .stop = ct_seq_stop, 292 .show = ct_seq_show 293 }; 294 295 static int ct_open(struct inode *inode, struct file *file) 296 { 297 return seq_open_net(inode, file, &ct_seq_ops, 298 sizeof(struct ct_iter_state)); 299 } 300 301 static const struct file_operations ct_file_ops = { 302 .owner = THIS_MODULE, 303 .open = ct_open, 304 .read = seq_read, 305 .llseek = seq_lseek, 306 .release = seq_release_net, 307 }; 308 309 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos) 310 { 311 struct net *net = seq_file_net(seq); 312 int cpu; 313 314 if (*pos == 0) 315 return SEQ_START_TOKEN; 316 317 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { 318 if (!cpu_possible(cpu)) 319 continue; 320 *pos = cpu + 1; 321 return per_cpu_ptr(net->ct.stat, cpu); 322 } 323 324 return NULL; 325 } 326 327 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) 328 { 329 struct net *net = seq_file_net(seq); 330 int cpu; 331 332 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { 333 if (!cpu_possible(cpu)) 334 continue; 335 *pos = cpu + 1; 336 return per_cpu_ptr(net->ct.stat, cpu); 337 } 338 339 return NULL; 340 } 341 342 static void ct_cpu_seq_stop(struct seq_file *seq, void *v) 343 { 344 } 345 346 static int ct_cpu_seq_show(struct seq_file *seq, void *v) 347 { 348 struct net *net = seq_file_net(seq); 349 unsigned int nr_conntracks = atomic_read(&net->ct.count); 350 const struct ip_conntrack_stat *st = v; 351 352 if (v == SEQ_START_TOKEN) { 353 seq_puts(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"); 354 return 0; 355 } 356 357 seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x " 358 "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n", 359 nr_conntracks, 360 0, 361 st->found, 362 0, 363 st->invalid, 364 st->ignore, 365 0, 366 0, 367 st->insert, 368 st->insert_failed, 369 st->drop, 370 st->early_drop, 371 st->error, 372 373 st->expect_new, 374 st->expect_create, 375 st->expect_delete, 376 st->search_restart 377 ); 378 return 0; 379 } 380 381 static const struct seq_operations ct_cpu_seq_ops = { 382 .start = ct_cpu_seq_start, 383 .next = ct_cpu_seq_next, 384 .stop = ct_cpu_seq_stop, 385 .show = ct_cpu_seq_show, 386 }; 387 388 static int ct_cpu_seq_open(struct inode *inode, struct file *file) 389 { 390 return seq_open_net(inode, file, &ct_cpu_seq_ops, 391 sizeof(struct seq_net_private)); 392 } 393 394 static const struct file_operations ct_cpu_seq_fops = { 395 .owner = THIS_MODULE, 396 .open = ct_cpu_seq_open, 397 .read = seq_read, 398 .llseek = seq_lseek, 399 .release = seq_release_net, 400 }; 401 402 static int nf_conntrack_standalone_init_proc(struct net *net) 403 { 404 struct proc_dir_entry *pde; 405 kuid_t root_uid; 406 kgid_t root_gid; 407 408 pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops); 409 if (!pde) 410 goto out_nf_conntrack; 411 412 root_uid = make_kuid(net->user_ns, 0); 413 root_gid = make_kgid(net->user_ns, 0); 414 if (uid_valid(root_uid) && gid_valid(root_gid)) 415 proc_set_user(pde, root_uid, root_gid); 416 417 pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat, 418 &ct_cpu_seq_fops); 419 if (!pde) 420 goto out_stat_nf_conntrack; 421 return 0; 422 423 out_stat_nf_conntrack: 424 remove_proc_entry("nf_conntrack", net->proc_net); 425 out_nf_conntrack: 426 return -ENOMEM; 427 } 428 429 static void nf_conntrack_standalone_fini_proc(struct net *net) 430 { 431 remove_proc_entry("nf_conntrack", net->proc_net_stat); 432 remove_proc_entry("nf_conntrack", net->proc_net); 433 } 434 #else 435 static int nf_conntrack_standalone_init_proc(struct net *net) 436 { 437 return 0; 438 } 439 440 static void nf_conntrack_standalone_fini_proc(struct net *net) 441 { 442 } 443 #endif /* CONFIG_NF_CONNTRACK_PROCFS */ 444 445 /* Sysctl support */ 446 447 #ifdef CONFIG_SYSCTL 448 /* Log invalid packets of a given protocol */ 449 static int log_invalid_proto_min __read_mostly; 450 static int log_invalid_proto_max __read_mostly = 255; 451 452 /* size the user *wants to set */ 453 static unsigned int nf_conntrack_htable_size_user __read_mostly; 454 455 extern unsigned int nf_conntrack_default_on; 456 unsigned int nf_conntrack_default_on __read_mostly = 1; 457 458 static int 459 nf_conntrack_hash_sysctl(struct ctl_table *table, int write, 460 void __user *buffer, size_t *lenp, loff_t *ppos) 461 { 462 int ret; 463 464 ret = proc_dointvec(table, write, buffer, lenp, ppos); 465 if (ret < 0 || !write) 466 return ret; 467 468 /* update ret, we might not be able to satisfy request */ 469 ret = nf_conntrack_hash_resize(nf_conntrack_htable_size_user); 470 471 /* update it to the actual value used by conntrack */ 472 nf_conntrack_htable_size_user = nf_conntrack_htable_size; 473 return ret; 474 } 475 476 static struct ctl_table_header *nf_ct_netfilter_header; 477 478 static struct ctl_table nf_ct_sysctl_table[] = { 479 { 480 .procname = "nf_conntrack_max", 481 .data = &nf_conntrack_max, 482 .maxlen = sizeof(int), 483 .mode = 0644, 484 .proc_handler = proc_dointvec, 485 }, 486 { 487 .procname = "nf_conntrack_count", 488 .data = &init_net.ct.count, 489 .maxlen = sizeof(int), 490 .mode = 0444, 491 .proc_handler = proc_dointvec, 492 }, 493 { 494 .procname = "nf_conntrack_buckets", 495 .data = &nf_conntrack_htable_size_user, 496 .maxlen = sizeof(unsigned int), 497 .mode = 0644, 498 .proc_handler = nf_conntrack_hash_sysctl, 499 }, 500 { 501 .procname = "nf_conntrack_checksum", 502 .data = &init_net.ct.sysctl_checksum, 503 .maxlen = sizeof(unsigned int), 504 .mode = 0644, 505 .proc_handler = proc_dointvec, 506 }, 507 { 508 .procname = "nf_conntrack_log_invalid", 509 .data = &init_net.ct.sysctl_log_invalid, 510 .maxlen = sizeof(unsigned int), 511 .mode = 0644, 512 .proc_handler = proc_dointvec_minmax, 513 .extra1 = &log_invalid_proto_min, 514 .extra2 = &log_invalid_proto_max, 515 }, 516 { 517 .procname = "nf_conntrack_expect_max", 518 .data = &nf_ct_expect_max, 519 .maxlen = sizeof(int), 520 .mode = 0644, 521 .proc_handler = proc_dointvec, 522 }, 523 { 524 .procname = "nf_conntrack_default_on", 525 .data = &nf_conntrack_default_on, 526 .maxlen = sizeof(unsigned int), 527 .mode = 0644, 528 .proc_handler = proc_dointvec, 529 }, 530 { } 531 }; 532 533 static struct ctl_table nf_ct_netfilter_table[] = { 534 { 535 .procname = "nf_conntrack_max", 536 .data = &nf_conntrack_max, 537 .maxlen = sizeof(int), 538 .mode = 0644, 539 .proc_handler = proc_dointvec, 540 }, 541 { } 542 }; 543 544 static int nf_conntrack_standalone_init_sysctl(struct net *net) 545 { 546 struct ctl_table *table; 547 548 table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table), 549 GFP_KERNEL); 550 if (!table) 551 goto out_kmemdup; 552 553 table[1].data = &net->ct.count; 554 table[3].data = &net->ct.sysctl_checksum; 555 table[4].data = &net->ct.sysctl_log_invalid; 556 557 /* Don't export sysctls to unprivileged users */ 558 if (net->user_ns != &init_user_ns) 559 table[0].procname = NULL; 560 561 if (!net_eq(&init_net, net)) 562 table[2].mode = 0444; 563 564 net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table); 565 if (!net->ct.sysctl_header) 566 goto out_unregister_netfilter; 567 568 return 0; 569 570 out_unregister_netfilter: 571 kfree(table); 572 out_kmemdup: 573 return -ENOMEM; 574 } 575 576 static void nf_conntrack_standalone_fini_sysctl(struct net *net) 577 { 578 struct ctl_table *table; 579 580 table = net->ct.sysctl_header->ctl_table_arg; 581 unregister_net_sysctl_table(net->ct.sysctl_header); 582 kfree(table); 583 } 584 #else 585 static int nf_conntrack_standalone_init_sysctl(struct net *net) 586 { 587 return 0; 588 } 589 590 static void nf_conntrack_standalone_fini_sysctl(struct net *net) 591 { 592 } 593 #endif /* CONFIG_SYSCTL */ 594 595 static int nf_conntrack_pernet_init(struct net *net) 596 { 597 int ret; 598 599 ret = nf_conntrack_init_net(net); 600 if (ret < 0) 601 goto out_init; 602 603 ret = nf_conntrack_standalone_init_proc(net); 604 if (ret < 0) 605 goto out_proc; 606 607 net->ct.sysctl_checksum = 1; 608 net->ct.sysctl_log_invalid = 0; 609 ret = nf_conntrack_standalone_init_sysctl(net); 610 if (ret < 0) 611 goto out_sysctl; 612 613 return 0; 614 615 out_sysctl: 616 nf_conntrack_standalone_fini_proc(net); 617 out_proc: 618 nf_conntrack_cleanup_net(net); 619 out_init: 620 return ret; 621 } 622 623 static void nf_conntrack_pernet_exit(struct list_head *net_exit_list) 624 { 625 struct net *net; 626 627 list_for_each_entry(net, net_exit_list, exit_list) { 628 nf_conntrack_standalone_fini_sysctl(net); 629 nf_conntrack_standalone_fini_proc(net); 630 } 631 nf_conntrack_cleanup_net_list(net_exit_list); 632 } 633 634 static struct pernet_operations nf_conntrack_net_ops = { 635 .init = nf_conntrack_pernet_init, 636 .exit_batch = nf_conntrack_pernet_exit, 637 }; 638 639 static int __init nf_conntrack_standalone_init(void) 640 { 641 int ret = nf_conntrack_init_start(); 642 if (ret < 0) 643 goto out_start; 644 645 BUILD_BUG_ON(SKB_NFCT_PTRMASK != NFCT_PTRMASK); 646 BUILD_BUG_ON(NFCT_INFOMASK <= IP_CT_NUMBER); 647 648 #ifdef CONFIG_SYSCTL 649 nf_ct_netfilter_header = 650 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table); 651 if (!nf_ct_netfilter_header) { 652 pr_err("nf_conntrack: can't register to sysctl.\n"); 653 ret = -ENOMEM; 654 goto out_sysctl; 655 } 656 657 nf_conntrack_htable_size_user = nf_conntrack_htable_size; 658 #endif 659 660 ret = register_pernet_subsys(&nf_conntrack_net_ops); 661 if (ret < 0) 662 goto out_pernet; 663 664 nf_conntrack_init_end(); 665 return 0; 666 667 out_pernet: 668 #ifdef CONFIG_SYSCTL 669 unregister_net_sysctl_table(nf_ct_netfilter_header); 670 out_sysctl: 671 #endif 672 nf_conntrack_cleanup_end(); 673 out_start: 674 return ret; 675 } 676 677 static void __exit nf_conntrack_standalone_fini(void) 678 { 679 nf_conntrack_cleanup_start(); 680 unregister_pernet_subsys(&nf_conntrack_net_ops); 681 #ifdef CONFIG_SYSCTL 682 unregister_net_sysctl_table(nf_ct_netfilter_header); 683 #endif 684 nf_conntrack_cleanup_end(); 685 } 686 687 module_init(nf_conntrack_standalone_init); 688 module_exit(nf_conntrack_standalone_fini); 689 690 /* Some modules need us, but don't depend directly on any symbol. 691 They should call this. */ 692 void need_conntrack(void) 693 { 694 } 695 EXPORT_SYMBOL_GPL(need_conntrack); 696