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