1 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 2 3 #include <linux/workqueue.h> 4 #include <linux/rtnetlink.h> 5 #include <linux/cache.h> 6 #include <linux/slab.h> 7 #include <linux/list.h> 8 #include <linux/delay.h> 9 #include <linux/sched.h> 10 #include <linux/idr.h> 11 #include <linux/rculist.h> 12 #include <linux/nsproxy.h> 13 #include <linux/fs.h> 14 #include <linux/proc_ns.h> 15 #include <linux/file.h> 16 #include <linux/export.h> 17 #include <linux/user_namespace.h> 18 #include <linux/net_namespace.h> 19 #include <net/sock.h> 20 #include <net/netlink.h> 21 #include <net/net_namespace.h> 22 #include <net/netns/generic.h> 23 24 /* 25 * Our network namespace constructor/destructor lists 26 */ 27 28 static LIST_HEAD(pernet_list); 29 static struct list_head *first_device = &pernet_list; 30 DEFINE_MUTEX(net_mutex); 31 32 LIST_HEAD(net_namespace_list); 33 EXPORT_SYMBOL_GPL(net_namespace_list); 34 35 struct net init_net = { 36 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), 37 }; 38 EXPORT_SYMBOL(init_net); 39 40 static bool init_net_initialized; 41 42 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ 43 44 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; 45 46 static struct net_generic *net_alloc_generic(void) 47 { 48 struct net_generic *ng; 49 size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); 50 51 ng = kzalloc(generic_size, GFP_KERNEL); 52 if (ng) 53 ng->s.len = max_gen_ptrs; 54 55 return ng; 56 } 57 58 static int net_assign_generic(struct net *net, unsigned int id, void *data) 59 { 60 struct net_generic *ng, *old_ng; 61 62 BUG_ON(!mutex_is_locked(&net_mutex)); 63 BUG_ON(id == 0); 64 65 old_ng = rcu_dereference_protected(net->gen, 66 lockdep_is_held(&net_mutex)); 67 if (old_ng->s.len >= id) { 68 old_ng->ptr[id - 1] = data; 69 return 0; 70 } 71 72 ng = net_alloc_generic(); 73 if (ng == NULL) 74 return -ENOMEM; 75 76 /* 77 * Some synchronisation notes: 78 * 79 * The net_generic explores the net->gen array inside rcu 80 * read section. Besides once set the net->gen->ptr[x] 81 * pointer never changes (see rules in netns/generic.h). 82 * 83 * That said, we simply duplicate this array and schedule 84 * the old copy for kfree after a grace period. 85 */ 86 87 memcpy(&ng->ptr, &old_ng->ptr, old_ng->s.len * sizeof(void*)); 88 ng->ptr[id - 1] = data; 89 90 rcu_assign_pointer(net->gen, ng); 91 kfree_rcu(old_ng, s.rcu); 92 return 0; 93 } 94 95 static int ops_init(const struct pernet_operations *ops, struct net *net) 96 { 97 int err = -ENOMEM; 98 void *data = NULL; 99 100 if (ops->id && ops->size) { 101 data = kzalloc(ops->size, GFP_KERNEL); 102 if (!data) 103 goto out; 104 105 err = net_assign_generic(net, *ops->id, data); 106 if (err) 107 goto cleanup; 108 } 109 err = 0; 110 if (ops->init) 111 err = ops->init(net); 112 if (!err) 113 return 0; 114 115 cleanup: 116 kfree(data); 117 118 out: 119 return err; 120 } 121 122 static void ops_free(const struct pernet_operations *ops, struct net *net) 123 { 124 if (ops->id && ops->size) { 125 kfree(net_generic(net, *ops->id)); 126 } 127 } 128 129 static void ops_exit_list(const struct pernet_operations *ops, 130 struct list_head *net_exit_list) 131 { 132 struct net *net; 133 if (ops->exit) { 134 list_for_each_entry(net, net_exit_list, exit_list) 135 ops->exit(net); 136 } 137 if (ops->exit_batch) 138 ops->exit_batch(net_exit_list); 139 } 140 141 static void ops_free_list(const struct pernet_operations *ops, 142 struct list_head *net_exit_list) 143 { 144 struct net *net; 145 if (ops->size && ops->id) { 146 list_for_each_entry(net, net_exit_list, exit_list) 147 ops_free(ops, net); 148 } 149 } 150 151 /* should be called with nsid_lock held */ 152 static int alloc_netid(struct net *net, struct net *peer, int reqid) 153 { 154 int min = 0, max = 0; 155 156 if (reqid >= 0) { 157 min = reqid; 158 max = reqid + 1; 159 } 160 161 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC); 162 } 163 164 /* This function is used by idr_for_each(). If net is equal to peer, the 165 * function returns the id so that idr_for_each() stops. Because we cannot 166 * returns the id 0 (idr_for_each() will not stop), we return the magic value 167 * NET_ID_ZERO (-1) for it. 168 */ 169 #define NET_ID_ZERO -1 170 static int net_eq_idr(int id, void *net, void *peer) 171 { 172 if (net_eq(net, peer)) 173 return id ? : NET_ID_ZERO; 174 return 0; 175 } 176 177 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc 178 * is set to true, thus the caller knows that the new id must be notified via 179 * rtnl. 180 */ 181 static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc) 182 { 183 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); 184 bool alloc_it = *alloc; 185 186 *alloc = false; 187 188 /* Magic value for id 0. */ 189 if (id == NET_ID_ZERO) 190 return 0; 191 if (id > 0) 192 return id; 193 194 if (alloc_it) { 195 id = alloc_netid(net, peer, -1); 196 *alloc = true; 197 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; 198 } 199 200 return NETNSA_NSID_NOT_ASSIGNED; 201 } 202 203 /* should be called with nsid_lock held */ 204 static int __peernet2id(struct net *net, struct net *peer) 205 { 206 bool no = false; 207 208 return __peernet2id_alloc(net, peer, &no); 209 } 210 211 static void rtnl_net_notifyid(struct net *net, int cmd, int id); 212 /* This function returns the id of a peer netns. If no id is assigned, one will 213 * be allocated and returned. 214 */ 215 int peernet2id_alloc(struct net *net, struct net *peer) 216 { 217 unsigned long flags; 218 bool alloc; 219 int id; 220 221 if (atomic_read(&net->count) == 0) 222 return NETNSA_NSID_NOT_ASSIGNED; 223 spin_lock_irqsave(&net->nsid_lock, flags); 224 alloc = atomic_read(&peer->count) == 0 ? false : true; 225 id = __peernet2id_alloc(net, peer, &alloc); 226 spin_unlock_irqrestore(&net->nsid_lock, flags); 227 if (alloc && id >= 0) 228 rtnl_net_notifyid(net, RTM_NEWNSID, id); 229 return id; 230 } 231 232 /* This function returns, if assigned, the id of a peer netns. */ 233 int peernet2id(struct net *net, struct net *peer) 234 { 235 unsigned long flags; 236 int id; 237 238 spin_lock_irqsave(&net->nsid_lock, flags); 239 id = __peernet2id(net, peer); 240 spin_unlock_irqrestore(&net->nsid_lock, flags); 241 return id; 242 } 243 EXPORT_SYMBOL(peernet2id); 244 245 /* This function returns true is the peer netns has an id assigned into the 246 * current netns. 247 */ 248 bool peernet_has_id(struct net *net, struct net *peer) 249 { 250 return peernet2id(net, peer) >= 0; 251 } 252 253 struct net *get_net_ns_by_id(struct net *net, int id) 254 { 255 unsigned long flags; 256 struct net *peer; 257 258 if (id < 0) 259 return NULL; 260 261 rcu_read_lock(); 262 spin_lock_irqsave(&net->nsid_lock, flags); 263 peer = idr_find(&net->netns_ids, id); 264 if (peer) 265 get_net(peer); 266 spin_unlock_irqrestore(&net->nsid_lock, flags); 267 rcu_read_unlock(); 268 269 return peer; 270 } 271 272 /* 273 * setup_net runs the initializers for the network namespace object. 274 */ 275 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) 276 { 277 /* Must be called with net_mutex held */ 278 const struct pernet_operations *ops, *saved_ops; 279 int error = 0; 280 LIST_HEAD(net_exit_list); 281 282 atomic_set(&net->count, 1); 283 atomic_set(&net->passive, 1); 284 net->dev_base_seq = 1; 285 net->user_ns = user_ns; 286 idr_init(&net->netns_ids); 287 spin_lock_init(&net->nsid_lock); 288 289 list_for_each_entry(ops, &pernet_list, list) { 290 error = ops_init(ops, net); 291 if (error < 0) 292 goto out_undo; 293 } 294 out: 295 return error; 296 297 out_undo: 298 /* Walk through the list backwards calling the exit functions 299 * for the pernet modules whose init functions did not fail. 300 */ 301 list_add(&net->exit_list, &net_exit_list); 302 saved_ops = ops; 303 list_for_each_entry_continue_reverse(ops, &pernet_list, list) 304 ops_exit_list(ops, &net_exit_list); 305 306 ops = saved_ops; 307 list_for_each_entry_continue_reverse(ops, &pernet_list, list) 308 ops_free_list(ops, &net_exit_list); 309 310 rcu_barrier(); 311 goto out; 312 } 313 314 315 #ifdef CONFIG_NET_NS 316 static struct ucounts *inc_net_namespaces(struct user_namespace *ns) 317 { 318 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES); 319 } 320 321 static void dec_net_namespaces(struct ucounts *ucounts) 322 { 323 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES); 324 } 325 326 static struct kmem_cache *net_cachep; 327 static struct workqueue_struct *netns_wq; 328 329 static struct net *net_alloc(void) 330 { 331 struct net *net = NULL; 332 struct net_generic *ng; 333 334 ng = net_alloc_generic(); 335 if (!ng) 336 goto out; 337 338 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); 339 if (!net) 340 goto out_free; 341 342 rcu_assign_pointer(net->gen, ng); 343 out: 344 return net; 345 346 out_free: 347 kfree(ng); 348 goto out; 349 } 350 351 static void net_free(struct net *net) 352 { 353 kfree(rcu_access_pointer(net->gen)); 354 kmem_cache_free(net_cachep, net); 355 } 356 357 void net_drop_ns(void *p) 358 { 359 struct net *ns = p; 360 if (ns && atomic_dec_and_test(&ns->passive)) 361 net_free(ns); 362 } 363 364 struct net *copy_net_ns(unsigned long flags, 365 struct user_namespace *user_ns, struct net *old_net) 366 { 367 struct ucounts *ucounts; 368 struct net *net; 369 int rv; 370 371 if (!(flags & CLONE_NEWNET)) 372 return get_net(old_net); 373 374 ucounts = inc_net_namespaces(user_ns); 375 if (!ucounts) 376 return ERR_PTR(-ENOSPC); 377 378 net = net_alloc(); 379 if (!net) { 380 dec_net_namespaces(ucounts); 381 return ERR_PTR(-ENOMEM); 382 } 383 384 get_user_ns(user_ns); 385 386 rv = mutex_lock_killable(&net_mutex); 387 if (rv < 0) { 388 net_free(net); 389 dec_net_namespaces(ucounts); 390 put_user_ns(user_ns); 391 return ERR_PTR(rv); 392 } 393 394 net->ucounts = ucounts; 395 rv = setup_net(net, user_ns); 396 if (rv == 0) { 397 rtnl_lock(); 398 list_add_tail_rcu(&net->list, &net_namespace_list); 399 rtnl_unlock(); 400 } 401 mutex_unlock(&net_mutex); 402 if (rv < 0) { 403 dec_net_namespaces(ucounts); 404 put_user_ns(user_ns); 405 net_drop_ns(net); 406 return ERR_PTR(rv); 407 } 408 return net; 409 } 410 411 static DEFINE_SPINLOCK(cleanup_list_lock); 412 static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ 413 414 static void cleanup_net(struct work_struct *work) 415 { 416 const struct pernet_operations *ops; 417 struct net *net, *tmp; 418 struct list_head net_kill_list; 419 LIST_HEAD(net_exit_list); 420 421 /* Atomically snapshot the list of namespaces to cleanup */ 422 spin_lock_irq(&cleanup_list_lock); 423 list_replace_init(&cleanup_list, &net_kill_list); 424 spin_unlock_irq(&cleanup_list_lock); 425 426 mutex_lock(&net_mutex); 427 428 /* Don't let anyone else find us. */ 429 rtnl_lock(); 430 list_for_each_entry(net, &net_kill_list, cleanup_list) { 431 list_del_rcu(&net->list); 432 list_add_tail(&net->exit_list, &net_exit_list); 433 for_each_net(tmp) { 434 int id; 435 436 spin_lock_irq(&tmp->nsid_lock); 437 id = __peernet2id(tmp, net); 438 if (id >= 0) 439 idr_remove(&tmp->netns_ids, id); 440 spin_unlock_irq(&tmp->nsid_lock); 441 if (id >= 0) 442 rtnl_net_notifyid(tmp, RTM_DELNSID, id); 443 } 444 spin_lock_irq(&net->nsid_lock); 445 idr_destroy(&net->netns_ids); 446 spin_unlock_irq(&net->nsid_lock); 447 448 } 449 rtnl_unlock(); 450 451 /* 452 * Another CPU might be rcu-iterating the list, wait for it. 453 * This needs to be before calling the exit() notifiers, so 454 * the rcu_barrier() below isn't sufficient alone. 455 */ 456 synchronize_rcu(); 457 458 /* Run all of the network namespace exit methods */ 459 list_for_each_entry_reverse(ops, &pernet_list, list) 460 ops_exit_list(ops, &net_exit_list); 461 462 /* Free the net generic variables */ 463 list_for_each_entry_reverse(ops, &pernet_list, list) 464 ops_free_list(ops, &net_exit_list); 465 466 mutex_unlock(&net_mutex); 467 468 /* Ensure there are no outstanding rcu callbacks using this 469 * network namespace. 470 */ 471 rcu_barrier(); 472 473 /* Finally it is safe to free my network namespace structure */ 474 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { 475 list_del_init(&net->exit_list); 476 dec_net_namespaces(net->ucounts); 477 put_user_ns(net->user_ns); 478 net_drop_ns(net); 479 } 480 } 481 static DECLARE_WORK(net_cleanup_work, cleanup_net); 482 483 void __put_net(struct net *net) 484 { 485 /* Cleanup the network namespace in process context */ 486 unsigned long flags; 487 488 spin_lock_irqsave(&cleanup_list_lock, flags); 489 list_add(&net->cleanup_list, &cleanup_list); 490 spin_unlock_irqrestore(&cleanup_list_lock, flags); 491 492 queue_work(netns_wq, &net_cleanup_work); 493 } 494 EXPORT_SYMBOL_GPL(__put_net); 495 496 struct net *get_net_ns_by_fd(int fd) 497 { 498 struct file *file; 499 struct ns_common *ns; 500 struct net *net; 501 502 file = proc_ns_fget(fd); 503 if (IS_ERR(file)) 504 return ERR_CAST(file); 505 506 ns = get_proc_ns(file_inode(file)); 507 if (ns->ops == &netns_operations) 508 net = get_net(container_of(ns, struct net, ns)); 509 else 510 net = ERR_PTR(-EINVAL); 511 512 fput(file); 513 return net; 514 } 515 516 #else 517 struct net *get_net_ns_by_fd(int fd) 518 { 519 return ERR_PTR(-EINVAL); 520 } 521 #endif 522 EXPORT_SYMBOL_GPL(get_net_ns_by_fd); 523 524 struct net *get_net_ns_by_pid(pid_t pid) 525 { 526 struct task_struct *tsk; 527 struct net *net; 528 529 /* Lookup the network namespace */ 530 net = ERR_PTR(-ESRCH); 531 rcu_read_lock(); 532 tsk = find_task_by_vpid(pid); 533 if (tsk) { 534 struct nsproxy *nsproxy; 535 task_lock(tsk); 536 nsproxy = tsk->nsproxy; 537 if (nsproxy) 538 net = get_net(nsproxy->net_ns); 539 task_unlock(tsk); 540 } 541 rcu_read_unlock(); 542 return net; 543 } 544 EXPORT_SYMBOL_GPL(get_net_ns_by_pid); 545 546 static __net_init int net_ns_net_init(struct net *net) 547 { 548 #ifdef CONFIG_NET_NS 549 net->ns.ops = &netns_operations; 550 #endif 551 return ns_alloc_inum(&net->ns); 552 } 553 554 static __net_exit void net_ns_net_exit(struct net *net) 555 { 556 ns_free_inum(&net->ns); 557 } 558 559 static struct pernet_operations __net_initdata net_ns_ops = { 560 .init = net_ns_net_init, 561 .exit = net_ns_net_exit, 562 }; 563 564 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { 565 [NETNSA_NONE] = { .type = NLA_UNSPEC }, 566 [NETNSA_NSID] = { .type = NLA_S32 }, 567 [NETNSA_PID] = { .type = NLA_U32 }, 568 [NETNSA_FD] = { .type = NLA_U32 }, 569 }; 570 571 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) 572 { 573 struct net *net = sock_net(skb->sk); 574 struct nlattr *tb[NETNSA_MAX + 1]; 575 unsigned long flags; 576 struct net *peer; 577 int nsid, err; 578 579 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, 580 rtnl_net_policy); 581 if (err < 0) 582 return err; 583 if (!tb[NETNSA_NSID]) 584 return -EINVAL; 585 nsid = nla_get_s32(tb[NETNSA_NSID]); 586 587 if (tb[NETNSA_PID]) 588 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); 589 else if (tb[NETNSA_FD]) 590 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); 591 else 592 return -EINVAL; 593 if (IS_ERR(peer)) 594 return PTR_ERR(peer); 595 596 spin_lock_irqsave(&net->nsid_lock, flags); 597 if (__peernet2id(net, peer) >= 0) { 598 spin_unlock_irqrestore(&net->nsid_lock, flags); 599 err = -EEXIST; 600 goto out; 601 } 602 603 err = alloc_netid(net, peer, nsid); 604 spin_unlock_irqrestore(&net->nsid_lock, flags); 605 if (err >= 0) { 606 rtnl_net_notifyid(net, RTM_NEWNSID, err); 607 err = 0; 608 } 609 out: 610 put_net(peer); 611 return err; 612 } 613 614 static int rtnl_net_get_size(void) 615 { 616 return NLMSG_ALIGN(sizeof(struct rtgenmsg)) 617 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ 618 ; 619 } 620 621 static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, 622 int cmd, struct net *net, int nsid) 623 { 624 struct nlmsghdr *nlh; 625 struct rtgenmsg *rth; 626 627 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); 628 if (!nlh) 629 return -EMSGSIZE; 630 631 rth = nlmsg_data(nlh); 632 rth->rtgen_family = AF_UNSPEC; 633 634 if (nla_put_s32(skb, NETNSA_NSID, nsid)) 635 goto nla_put_failure; 636 637 nlmsg_end(skb, nlh); 638 return 0; 639 640 nla_put_failure: 641 nlmsg_cancel(skb, nlh); 642 return -EMSGSIZE; 643 } 644 645 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) 646 { 647 struct net *net = sock_net(skb->sk); 648 struct nlattr *tb[NETNSA_MAX + 1]; 649 struct sk_buff *msg; 650 struct net *peer; 651 int err, id; 652 653 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, 654 rtnl_net_policy); 655 if (err < 0) 656 return err; 657 if (tb[NETNSA_PID]) 658 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); 659 else if (tb[NETNSA_FD]) 660 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); 661 else 662 return -EINVAL; 663 664 if (IS_ERR(peer)) 665 return PTR_ERR(peer); 666 667 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); 668 if (!msg) { 669 err = -ENOMEM; 670 goto out; 671 } 672 673 id = peernet2id(net, peer); 674 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 675 RTM_NEWNSID, net, id); 676 if (err < 0) 677 goto err_out; 678 679 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); 680 goto out; 681 682 err_out: 683 nlmsg_free(msg); 684 out: 685 put_net(peer); 686 return err; 687 } 688 689 struct rtnl_net_dump_cb { 690 struct net *net; 691 struct sk_buff *skb; 692 struct netlink_callback *cb; 693 int idx; 694 int s_idx; 695 }; 696 697 static int rtnl_net_dumpid_one(int id, void *peer, void *data) 698 { 699 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; 700 int ret; 701 702 if (net_cb->idx < net_cb->s_idx) 703 goto cont; 704 705 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, 706 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, 707 RTM_NEWNSID, net_cb->net, id); 708 if (ret < 0) 709 return ret; 710 711 cont: 712 net_cb->idx++; 713 return 0; 714 } 715 716 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) 717 { 718 struct net *net = sock_net(skb->sk); 719 struct rtnl_net_dump_cb net_cb = { 720 .net = net, 721 .skb = skb, 722 .cb = cb, 723 .idx = 0, 724 .s_idx = cb->args[0], 725 }; 726 unsigned long flags; 727 728 spin_lock_irqsave(&net->nsid_lock, flags); 729 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); 730 spin_unlock_irqrestore(&net->nsid_lock, flags); 731 732 cb->args[0] = net_cb.idx; 733 return skb->len; 734 } 735 736 static void rtnl_net_notifyid(struct net *net, int cmd, int id) 737 { 738 struct sk_buff *msg; 739 int err = -ENOMEM; 740 741 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); 742 if (!msg) 743 goto out; 744 745 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); 746 if (err < 0) 747 goto err_out; 748 749 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); 750 return; 751 752 err_out: 753 nlmsg_free(msg); 754 out: 755 rtnl_set_sk_err(net, RTNLGRP_NSID, err); 756 } 757 758 static int __init net_ns_init(void) 759 { 760 struct net_generic *ng; 761 762 #ifdef CONFIG_NET_NS 763 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), 764 SMP_CACHE_BYTES, 765 SLAB_PANIC, NULL); 766 767 /* Create workqueue for cleanup */ 768 netns_wq = create_singlethread_workqueue("netns"); 769 if (!netns_wq) 770 panic("Could not create netns workq"); 771 #endif 772 773 ng = net_alloc_generic(); 774 if (!ng) 775 panic("Could not allocate generic netns"); 776 777 rcu_assign_pointer(init_net.gen, ng); 778 779 mutex_lock(&net_mutex); 780 if (setup_net(&init_net, &init_user_ns)) 781 panic("Could not setup the initial network namespace"); 782 783 init_net_initialized = true; 784 785 rtnl_lock(); 786 list_add_tail_rcu(&init_net.list, &net_namespace_list); 787 rtnl_unlock(); 788 789 mutex_unlock(&net_mutex); 790 791 register_pernet_subsys(&net_ns_ops); 792 793 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); 794 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, 795 NULL); 796 797 return 0; 798 } 799 800 pure_initcall(net_ns_init); 801 802 #ifdef CONFIG_NET_NS 803 static int __register_pernet_operations(struct list_head *list, 804 struct pernet_operations *ops) 805 { 806 struct net *net; 807 int error; 808 LIST_HEAD(net_exit_list); 809 810 list_add_tail(&ops->list, list); 811 if (ops->init || (ops->id && ops->size)) { 812 for_each_net(net) { 813 error = ops_init(ops, net); 814 if (error) 815 goto out_undo; 816 list_add_tail(&net->exit_list, &net_exit_list); 817 } 818 } 819 return 0; 820 821 out_undo: 822 /* If I have an error cleanup all namespaces I initialized */ 823 list_del(&ops->list); 824 ops_exit_list(ops, &net_exit_list); 825 ops_free_list(ops, &net_exit_list); 826 return error; 827 } 828 829 static void __unregister_pernet_operations(struct pernet_operations *ops) 830 { 831 struct net *net; 832 LIST_HEAD(net_exit_list); 833 834 list_del(&ops->list); 835 for_each_net(net) 836 list_add_tail(&net->exit_list, &net_exit_list); 837 ops_exit_list(ops, &net_exit_list); 838 ops_free_list(ops, &net_exit_list); 839 } 840 841 #else 842 843 static int __register_pernet_operations(struct list_head *list, 844 struct pernet_operations *ops) 845 { 846 if (!init_net_initialized) { 847 list_add_tail(&ops->list, list); 848 return 0; 849 } 850 851 return ops_init(ops, &init_net); 852 } 853 854 static void __unregister_pernet_operations(struct pernet_operations *ops) 855 { 856 if (!init_net_initialized) { 857 list_del(&ops->list); 858 } else { 859 LIST_HEAD(net_exit_list); 860 list_add(&init_net.exit_list, &net_exit_list); 861 ops_exit_list(ops, &net_exit_list); 862 ops_free_list(ops, &net_exit_list); 863 } 864 } 865 866 #endif /* CONFIG_NET_NS */ 867 868 static DEFINE_IDA(net_generic_ids); 869 870 static int register_pernet_operations(struct list_head *list, 871 struct pernet_operations *ops) 872 { 873 int error; 874 875 if (ops->id) { 876 again: 877 error = ida_get_new_above(&net_generic_ids, 1, ops->id); 878 if (error < 0) { 879 if (error == -EAGAIN) { 880 ida_pre_get(&net_generic_ids, GFP_KERNEL); 881 goto again; 882 } 883 return error; 884 } 885 max_gen_ptrs = max(max_gen_ptrs, *ops->id); 886 } 887 error = __register_pernet_operations(list, ops); 888 if (error) { 889 rcu_barrier(); 890 if (ops->id) 891 ida_remove(&net_generic_ids, *ops->id); 892 } 893 894 return error; 895 } 896 897 static void unregister_pernet_operations(struct pernet_operations *ops) 898 { 899 900 __unregister_pernet_operations(ops); 901 rcu_barrier(); 902 if (ops->id) 903 ida_remove(&net_generic_ids, *ops->id); 904 } 905 906 /** 907 * register_pernet_subsys - register a network namespace subsystem 908 * @ops: pernet operations structure for the subsystem 909 * 910 * Register a subsystem which has init and exit functions 911 * that are called when network namespaces are created and 912 * destroyed respectively. 913 * 914 * When registered all network namespace init functions are 915 * called for every existing network namespace. Allowing kernel 916 * modules to have a race free view of the set of network namespaces. 917 * 918 * When a new network namespace is created all of the init 919 * methods are called in the order in which they were registered. 920 * 921 * When a network namespace is destroyed all of the exit methods 922 * are called in the reverse of the order with which they were 923 * registered. 924 */ 925 int register_pernet_subsys(struct pernet_operations *ops) 926 { 927 int error; 928 mutex_lock(&net_mutex); 929 error = register_pernet_operations(first_device, ops); 930 mutex_unlock(&net_mutex); 931 return error; 932 } 933 EXPORT_SYMBOL_GPL(register_pernet_subsys); 934 935 /** 936 * unregister_pernet_subsys - unregister a network namespace subsystem 937 * @ops: pernet operations structure to manipulate 938 * 939 * Remove the pernet operations structure from the list to be 940 * used when network namespaces are created or destroyed. In 941 * addition run the exit method for all existing network 942 * namespaces. 943 */ 944 void unregister_pernet_subsys(struct pernet_operations *ops) 945 { 946 mutex_lock(&net_mutex); 947 unregister_pernet_operations(ops); 948 mutex_unlock(&net_mutex); 949 } 950 EXPORT_SYMBOL_GPL(unregister_pernet_subsys); 951 952 /** 953 * register_pernet_device - register a network namespace device 954 * @ops: pernet operations structure for the subsystem 955 * 956 * Register a device which has init and exit functions 957 * that are called when network namespaces are created and 958 * destroyed respectively. 959 * 960 * When registered all network namespace init functions are 961 * called for every existing network namespace. Allowing kernel 962 * modules to have a race free view of the set of network namespaces. 963 * 964 * When a new network namespace is created all of the init 965 * methods are called in the order in which they were registered. 966 * 967 * When a network namespace is destroyed all of the exit methods 968 * are called in the reverse of the order with which they were 969 * registered. 970 */ 971 int register_pernet_device(struct pernet_operations *ops) 972 { 973 int error; 974 mutex_lock(&net_mutex); 975 error = register_pernet_operations(&pernet_list, ops); 976 if (!error && (first_device == &pernet_list)) 977 first_device = &ops->list; 978 mutex_unlock(&net_mutex); 979 return error; 980 } 981 EXPORT_SYMBOL_GPL(register_pernet_device); 982 983 /** 984 * unregister_pernet_device - unregister a network namespace netdevice 985 * @ops: pernet operations structure to manipulate 986 * 987 * Remove the pernet operations structure from the list to be 988 * used when network namespaces are created or destroyed. In 989 * addition run the exit method for all existing network 990 * namespaces. 991 */ 992 void unregister_pernet_device(struct pernet_operations *ops) 993 { 994 mutex_lock(&net_mutex); 995 if (&ops->list == first_device) 996 first_device = first_device->next; 997 unregister_pernet_operations(ops); 998 mutex_unlock(&net_mutex); 999 } 1000 EXPORT_SYMBOL_GPL(unregister_pernet_device); 1001 1002 #ifdef CONFIG_NET_NS 1003 static struct ns_common *netns_get(struct task_struct *task) 1004 { 1005 struct net *net = NULL; 1006 struct nsproxy *nsproxy; 1007 1008 task_lock(task); 1009 nsproxy = task->nsproxy; 1010 if (nsproxy) 1011 net = get_net(nsproxy->net_ns); 1012 task_unlock(task); 1013 1014 return net ? &net->ns : NULL; 1015 } 1016 1017 static inline struct net *to_net_ns(struct ns_common *ns) 1018 { 1019 return container_of(ns, struct net, ns); 1020 } 1021 1022 static void netns_put(struct ns_common *ns) 1023 { 1024 put_net(to_net_ns(ns)); 1025 } 1026 1027 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) 1028 { 1029 struct net *net = to_net_ns(ns); 1030 1031 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || 1032 !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) 1033 return -EPERM; 1034 1035 put_net(nsproxy->net_ns); 1036 nsproxy->net_ns = get_net(net); 1037 return 0; 1038 } 1039 1040 static struct user_namespace *netns_owner(struct ns_common *ns) 1041 { 1042 return to_net_ns(ns)->user_ns; 1043 } 1044 1045 const struct proc_ns_operations netns_operations = { 1046 .name = "net", 1047 .type = CLONE_NEWNET, 1048 .get = netns_get, 1049 .put = netns_put, 1050 .install = netns_install, 1051 .owner = netns_owner, 1052 }; 1053 #endif 1054