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