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 ucounts *inc_net_namespaces(struct user_namespace *ns) 313 { 314 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES); 315 } 316 317 static void dec_net_namespaces(struct ucounts *ucounts) 318 { 319 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES); 320 } 321 322 static struct kmem_cache *net_cachep; 323 static struct workqueue_struct *netns_wq; 324 325 static struct net *net_alloc(void) 326 { 327 struct net *net = NULL; 328 struct net_generic *ng; 329 330 ng = net_alloc_generic(); 331 if (!ng) 332 goto out; 333 334 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); 335 if (!net) 336 goto out_free; 337 338 rcu_assign_pointer(net->gen, ng); 339 out: 340 return net; 341 342 out_free: 343 kfree(ng); 344 goto out; 345 } 346 347 static void net_free(struct net *net) 348 { 349 kfree(rcu_access_pointer(net->gen)); 350 kmem_cache_free(net_cachep, net); 351 } 352 353 void net_drop_ns(void *p) 354 { 355 struct net *ns = p; 356 if (ns && atomic_dec_and_test(&ns->passive)) 357 net_free(ns); 358 } 359 360 struct net *copy_net_ns(unsigned long flags, 361 struct user_namespace *user_ns, struct net *old_net) 362 { 363 struct ucounts *ucounts; 364 struct net *net; 365 int rv; 366 367 if (!(flags & CLONE_NEWNET)) 368 return get_net(old_net); 369 370 ucounts = inc_net_namespaces(user_ns); 371 if (!ucounts) 372 return ERR_PTR(-ENOSPC); 373 374 net = net_alloc(); 375 if (!net) { 376 dec_net_namespaces(ucounts); 377 return ERR_PTR(-ENOMEM); 378 } 379 380 get_user_ns(user_ns); 381 382 rv = mutex_lock_killable(&net_mutex); 383 if (rv < 0) { 384 net_free(net); 385 dec_net_namespaces(ucounts); 386 put_user_ns(user_ns); 387 return ERR_PTR(rv); 388 } 389 390 net->ucounts = ucounts; 391 rv = setup_net(net, user_ns); 392 if (rv == 0) { 393 rtnl_lock(); 394 list_add_tail_rcu(&net->list, &net_namespace_list); 395 rtnl_unlock(); 396 } 397 mutex_unlock(&net_mutex); 398 if (rv < 0) { 399 dec_net_namespaces(ucounts); 400 put_user_ns(user_ns); 401 net_drop_ns(net); 402 return ERR_PTR(rv); 403 } 404 return net; 405 } 406 407 static DEFINE_SPINLOCK(cleanup_list_lock); 408 static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ 409 410 static void cleanup_net(struct work_struct *work) 411 { 412 const struct pernet_operations *ops; 413 struct net *net, *tmp; 414 struct list_head net_kill_list; 415 LIST_HEAD(net_exit_list); 416 417 /* Atomically snapshot the list of namespaces to cleanup */ 418 spin_lock_irq(&cleanup_list_lock); 419 list_replace_init(&cleanup_list, &net_kill_list); 420 spin_unlock_irq(&cleanup_list_lock); 421 422 mutex_lock(&net_mutex); 423 424 /* Don't let anyone else find us. */ 425 rtnl_lock(); 426 list_for_each_entry(net, &net_kill_list, cleanup_list) { 427 list_del_rcu(&net->list); 428 list_add_tail(&net->exit_list, &net_exit_list); 429 for_each_net(tmp) { 430 int id; 431 432 spin_lock_bh(&tmp->nsid_lock); 433 id = __peernet2id(tmp, net); 434 if (id >= 0) 435 idr_remove(&tmp->netns_ids, id); 436 spin_unlock_bh(&tmp->nsid_lock); 437 if (id >= 0) 438 rtnl_net_notifyid(tmp, RTM_DELNSID, id); 439 } 440 spin_lock_bh(&net->nsid_lock); 441 idr_destroy(&net->netns_ids); 442 spin_unlock_bh(&net->nsid_lock); 443 444 } 445 rtnl_unlock(); 446 447 /* 448 * Another CPU might be rcu-iterating the list, wait for it. 449 * This needs to be before calling the exit() notifiers, so 450 * the rcu_barrier() below isn't sufficient alone. 451 */ 452 synchronize_rcu(); 453 454 /* Run all of the network namespace exit methods */ 455 list_for_each_entry_reverse(ops, &pernet_list, list) 456 ops_exit_list(ops, &net_exit_list); 457 458 /* Free the net generic variables */ 459 list_for_each_entry_reverse(ops, &pernet_list, list) 460 ops_free_list(ops, &net_exit_list); 461 462 mutex_unlock(&net_mutex); 463 464 /* Ensure there are no outstanding rcu callbacks using this 465 * network namespace. 466 */ 467 rcu_barrier(); 468 469 /* Finally it is safe to free my network namespace structure */ 470 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { 471 list_del_init(&net->exit_list); 472 dec_net_namespaces(net->ucounts); 473 put_user_ns(net->user_ns); 474 net_drop_ns(net); 475 } 476 } 477 static DECLARE_WORK(net_cleanup_work, cleanup_net); 478 479 void __put_net(struct net *net) 480 { 481 /* Cleanup the network namespace in process context */ 482 unsigned long flags; 483 484 spin_lock_irqsave(&cleanup_list_lock, flags); 485 list_add(&net->cleanup_list, &cleanup_list); 486 spin_unlock_irqrestore(&cleanup_list_lock, flags); 487 488 queue_work(netns_wq, &net_cleanup_work); 489 } 490 EXPORT_SYMBOL_GPL(__put_net); 491 492 struct net *get_net_ns_by_fd(int fd) 493 { 494 struct file *file; 495 struct ns_common *ns; 496 struct net *net; 497 498 file = proc_ns_fget(fd); 499 if (IS_ERR(file)) 500 return ERR_CAST(file); 501 502 ns = get_proc_ns(file_inode(file)); 503 if (ns->ops == &netns_operations) 504 net = get_net(container_of(ns, struct net, ns)); 505 else 506 net = ERR_PTR(-EINVAL); 507 508 fput(file); 509 return net; 510 } 511 512 #else 513 struct net *get_net_ns_by_fd(int fd) 514 { 515 return ERR_PTR(-EINVAL); 516 } 517 #endif 518 EXPORT_SYMBOL_GPL(get_net_ns_by_fd); 519 520 struct net *get_net_ns_by_pid(pid_t pid) 521 { 522 struct task_struct *tsk; 523 struct net *net; 524 525 /* Lookup the network namespace */ 526 net = ERR_PTR(-ESRCH); 527 rcu_read_lock(); 528 tsk = find_task_by_vpid(pid); 529 if (tsk) { 530 struct nsproxy *nsproxy; 531 task_lock(tsk); 532 nsproxy = tsk->nsproxy; 533 if (nsproxy) 534 net = get_net(nsproxy->net_ns); 535 task_unlock(tsk); 536 } 537 rcu_read_unlock(); 538 return net; 539 } 540 EXPORT_SYMBOL_GPL(get_net_ns_by_pid); 541 542 static __net_init int net_ns_net_init(struct net *net) 543 { 544 #ifdef CONFIG_NET_NS 545 net->ns.ops = &netns_operations; 546 #endif 547 return ns_alloc_inum(&net->ns); 548 } 549 550 static __net_exit void net_ns_net_exit(struct net *net) 551 { 552 ns_free_inum(&net->ns); 553 } 554 555 static struct pernet_operations __net_initdata net_ns_ops = { 556 .init = net_ns_net_init, 557 .exit = net_ns_net_exit, 558 }; 559 560 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { 561 [NETNSA_NONE] = { .type = NLA_UNSPEC }, 562 [NETNSA_NSID] = { .type = NLA_S32 }, 563 [NETNSA_PID] = { .type = NLA_U32 }, 564 [NETNSA_FD] = { .type = NLA_U32 }, 565 }; 566 567 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) 568 { 569 struct net *net = sock_net(skb->sk); 570 struct nlattr *tb[NETNSA_MAX + 1]; 571 struct net *peer; 572 int nsid, err; 573 574 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, 575 rtnl_net_policy); 576 if (err < 0) 577 return err; 578 if (!tb[NETNSA_NSID]) 579 return -EINVAL; 580 nsid = nla_get_s32(tb[NETNSA_NSID]); 581 582 if (tb[NETNSA_PID]) 583 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); 584 else if (tb[NETNSA_FD]) 585 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); 586 else 587 return -EINVAL; 588 if (IS_ERR(peer)) 589 return PTR_ERR(peer); 590 591 spin_lock_bh(&net->nsid_lock); 592 if (__peernet2id(net, peer) >= 0) { 593 spin_unlock_bh(&net->nsid_lock); 594 err = -EEXIST; 595 goto out; 596 } 597 598 err = alloc_netid(net, peer, nsid); 599 spin_unlock_bh(&net->nsid_lock); 600 if (err >= 0) { 601 rtnl_net_notifyid(net, RTM_NEWNSID, err); 602 err = 0; 603 } 604 out: 605 put_net(peer); 606 return err; 607 } 608 609 static int rtnl_net_get_size(void) 610 { 611 return NLMSG_ALIGN(sizeof(struct rtgenmsg)) 612 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ 613 ; 614 } 615 616 static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, 617 int cmd, struct net *net, int nsid) 618 { 619 struct nlmsghdr *nlh; 620 struct rtgenmsg *rth; 621 622 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); 623 if (!nlh) 624 return -EMSGSIZE; 625 626 rth = nlmsg_data(nlh); 627 rth->rtgen_family = AF_UNSPEC; 628 629 if (nla_put_s32(skb, NETNSA_NSID, nsid)) 630 goto nla_put_failure; 631 632 nlmsg_end(skb, nlh); 633 return 0; 634 635 nla_put_failure: 636 nlmsg_cancel(skb, nlh); 637 return -EMSGSIZE; 638 } 639 640 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) 641 { 642 struct net *net = sock_net(skb->sk); 643 struct nlattr *tb[NETNSA_MAX + 1]; 644 struct sk_buff *msg; 645 struct net *peer; 646 int err, id; 647 648 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, 649 rtnl_net_policy); 650 if (err < 0) 651 return err; 652 if (tb[NETNSA_PID]) 653 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); 654 else if (tb[NETNSA_FD]) 655 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); 656 else 657 return -EINVAL; 658 659 if (IS_ERR(peer)) 660 return PTR_ERR(peer); 661 662 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); 663 if (!msg) { 664 err = -ENOMEM; 665 goto out; 666 } 667 668 id = peernet2id(net, peer); 669 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 670 RTM_NEWNSID, net, id); 671 if (err < 0) 672 goto err_out; 673 674 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); 675 goto out; 676 677 err_out: 678 nlmsg_free(msg); 679 out: 680 put_net(peer); 681 return err; 682 } 683 684 struct rtnl_net_dump_cb { 685 struct net *net; 686 struct sk_buff *skb; 687 struct netlink_callback *cb; 688 int idx; 689 int s_idx; 690 }; 691 692 static int rtnl_net_dumpid_one(int id, void *peer, void *data) 693 { 694 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; 695 int ret; 696 697 if (net_cb->idx < net_cb->s_idx) 698 goto cont; 699 700 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, 701 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, 702 RTM_NEWNSID, net_cb->net, id); 703 if (ret < 0) 704 return ret; 705 706 cont: 707 net_cb->idx++; 708 return 0; 709 } 710 711 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) 712 { 713 struct net *net = sock_net(skb->sk); 714 struct rtnl_net_dump_cb net_cb = { 715 .net = net, 716 .skb = skb, 717 .cb = cb, 718 .idx = 0, 719 .s_idx = cb->args[0], 720 }; 721 722 spin_lock_bh(&net->nsid_lock); 723 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); 724 spin_unlock_bh(&net->nsid_lock); 725 726 cb->args[0] = net_cb.idx; 727 return skb->len; 728 } 729 730 static void rtnl_net_notifyid(struct net *net, int cmd, int id) 731 { 732 struct sk_buff *msg; 733 int err = -ENOMEM; 734 735 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); 736 if (!msg) 737 goto out; 738 739 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); 740 if (err < 0) 741 goto err_out; 742 743 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); 744 return; 745 746 err_out: 747 nlmsg_free(msg); 748 out: 749 rtnl_set_sk_err(net, RTNLGRP_NSID, err); 750 } 751 752 static int __init net_ns_init(void) 753 { 754 struct net_generic *ng; 755 756 #ifdef CONFIG_NET_NS 757 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), 758 SMP_CACHE_BYTES, 759 SLAB_PANIC, NULL); 760 761 /* Create workqueue for cleanup */ 762 netns_wq = create_singlethread_workqueue("netns"); 763 if (!netns_wq) 764 panic("Could not create netns workq"); 765 #endif 766 767 ng = net_alloc_generic(); 768 if (!ng) 769 panic("Could not allocate generic netns"); 770 771 rcu_assign_pointer(init_net.gen, ng); 772 773 mutex_lock(&net_mutex); 774 if (setup_net(&init_net, &init_user_ns)) 775 panic("Could not setup the initial network namespace"); 776 777 init_net_initialized = true; 778 779 rtnl_lock(); 780 list_add_tail_rcu(&init_net.list, &net_namespace_list); 781 rtnl_unlock(); 782 783 mutex_unlock(&net_mutex); 784 785 register_pernet_subsys(&net_ns_ops); 786 787 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); 788 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, 789 NULL); 790 791 return 0; 792 } 793 794 pure_initcall(net_ns_init); 795 796 #ifdef CONFIG_NET_NS 797 static int __register_pernet_operations(struct list_head *list, 798 struct pernet_operations *ops) 799 { 800 struct net *net; 801 int error; 802 LIST_HEAD(net_exit_list); 803 804 list_add_tail(&ops->list, list); 805 if (ops->init || (ops->id && ops->size)) { 806 for_each_net(net) { 807 error = ops_init(ops, net); 808 if (error) 809 goto out_undo; 810 list_add_tail(&net->exit_list, &net_exit_list); 811 } 812 } 813 return 0; 814 815 out_undo: 816 /* If I have an error cleanup all namespaces I initialized */ 817 list_del(&ops->list); 818 ops_exit_list(ops, &net_exit_list); 819 ops_free_list(ops, &net_exit_list); 820 return error; 821 } 822 823 static void __unregister_pernet_operations(struct pernet_operations *ops) 824 { 825 struct net *net; 826 LIST_HEAD(net_exit_list); 827 828 list_del(&ops->list); 829 for_each_net(net) 830 list_add_tail(&net->exit_list, &net_exit_list); 831 ops_exit_list(ops, &net_exit_list); 832 ops_free_list(ops, &net_exit_list); 833 } 834 835 #else 836 837 static int __register_pernet_operations(struct list_head *list, 838 struct pernet_operations *ops) 839 { 840 if (!init_net_initialized) { 841 list_add_tail(&ops->list, list); 842 return 0; 843 } 844 845 return ops_init(ops, &init_net); 846 } 847 848 static void __unregister_pernet_operations(struct pernet_operations *ops) 849 { 850 if (!init_net_initialized) { 851 list_del(&ops->list); 852 } else { 853 LIST_HEAD(net_exit_list); 854 list_add(&init_net.exit_list, &net_exit_list); 855 ops_exit_list(ops, &net_exit_list); 856 ops_free_list(ops, &net_exit_list); 857 } 858 } 859 860 #endif /* CONFIG_NET_NS */ 861 862 static DEFINE_IDA(net_generic_ids); 863 864 static int register_pernet_operations(struct list_head *list, 865 struct pernet_operations *ops) 866 { 867 int error; 868 869 if (ops->id) { 870 again: 871 error = ida_get_new_above(&net_generic_ids, 1, ops->id); 872 if (error < 0) { 873 if (error == -EAGAIN) { 874 ida_pre_get(&net_generic_ids, GFP_KERNEL); 875 goto again; 876 } 877 return error; 878 } 879 max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); 880 } 881 error = __register_pernet_operations(list, ops); 882 if (error) { 883 rcu_barrier(); 884 if (ops->id) 885 ida_remove(&net_generic_ids, *ops->id); 886 } 887 888 return error; 889 } 890 891 static void unregister_pernet_operations(struct pernet_operations *ops) 892 { 893 894 __unregister_pernet_operations(ops); 895 rcu_barrier(); 896 if (ops->id) 897 ida_remove(&net_generic_ids, *ops->id); 898 } 899 900 /** 901 * register_pernet_subsys - register a network namespace subsystem 902 * @ops: pernet operations structure for the subsystem 903 * 904 * Register a subsystem which has init and exit functions 905 * that are called when network namespaces are created and 906 * destroyed respectively. 907 * 908 * When registered all network namespace init functions are 909 * called for every existing network namespace. Allowing kernel 910 * modules to have a race free view of the set of network namespaces. 911 * 912 * When a new network namespace is created all of the init 913 * methods are called in the order in which they were registered. 914 * 915 * When a network namespace is destroyed all of the exit methods 916 * are called in the reverse of the order with which they were 917 * registered. 918 */ 919 int register_pernet_subsys(struct pernet_operations *ops) 920 { 921 int error; 922 mutex_lock(&net_mutex); 923 error = register_pernet_operations(first_device, ops); 924 mutex_unlock(&net_mutex); 925 return error; 926 } 927 EXPORT_SYMBOL_GPL(register_pernet_subsys); 928 929 /** 930 * unregister_pernet_subsys - unregister a network namespace subsystem 931 * @ops: pernet operations structure to manipulate 932 * 933 * Remove the pernet operations structure from the list to be 934 * used when network namespaces are created or destroyed. In 935 * addition run the exit method for all existing network 936 * namespaces. 937 */ 938 void unregister_pernet_subsys(struct pernet_operations *ops) 939 { 940 mutex_lock(&net_mutex); 941 unregister_pernet_operations(ops); 942 mutex_unlock(&net_mutex); 943 } 944 EXPORT_SYMBOL_GPL(unregister_pernet_subsys); 945 946 /** 947 * register_pernet_device - register a network namespace device 948 * @ops: pernet operations structure for the subsystem 949 * 950 * Register a device which has init and exit functions 951 * that are called when network namespaces are created and 952 * destroyed respectively. 953 * 954 * When registered all network namespace init functions are 955 * called for every existing network namespace. Allowing kernel 956 * modules to have a race free view of the set of network namespaces. 957 * 958 * When a new network namespace is created all of the init 959 * methods are called in the order in which they were registered. 960 * 961 * When a network namespace is destroyed all of the exit methods 962 * are called in the reverse of the order with which they were 963 * registered. 964 */ 965 int register_pernet_device(struct pernet_operations *ops) 966 { 967 int error; 968 mutex_lock(&net_mutex); 969 error = register_pernet_operations(&pernet_list, ops); 970 if (!error && (first_device == &pernet_list)) 971 first_device = &ops->list; 972 mutex_unlock(&net_mutex); 973 return error; 974 } 975 EXPORT_SYMBOL_GPL(register_pernet_device); 976 977 /** 978 * unregister_pernet_device - unregister a network namespace netdevice 979 * @ops: pernet operations structure to manipulate 980 * 981 * Remove the pernet operations structure from the list to be 982 * used when network namespaces are created or destroyed. In 983 * addition run the exit method for all existing network 984 * namespaces. 985 */ 986 void unregister_pernet_device(struct pernet_operations *ops) 987 { 988 mutex_lock(&net_mutex); 989 if (&ops->list == first_device) 990 first_device = first_device->next; 991 unregister_pernet_operations(ops); 992 mutex_unlock(&net_mutex); 993 } 994 EXPORT_SYMBOL_GPL(unregister_pernet_device); 995 996 #ifdef CONFIG_NET_NS 997 static struct ns_common *netns_get(struct task_struct *task) 998 { 999 struct net *net = NULL; 1000 struct nsproxy *nsproxy; 1001 1002 task_lock(task); 1003 nsproxy = task->nsproxy; 1004 if (nsproxy) 1005 net = get_net(nsproxy->net_ns); 1006 task_unlock(task); 1007 1008 return net ? &net->ns : NULL; 1009 } 1010 1011 static inline struct net *to_net_ns(struct ns_common *ns) 1012 { 1013 return container_of(ns, struct net, ns); 1014 } 1015 1016 static void netns_put(struct ns_common *ns) 1017 { 1018 put_net(to_net_ns(ns)); 1019 } 1020 1021 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) 1022 { 1023 struct net *net = to_net_ns(ns); 1024 1025 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || 1026 !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) 1027 return -EPERM; 1028 1029 put_net(nsproxy->net_ns); 1030 nsproxy->net_ns = get_net(net); 1031 return 0; 1032 } 1033 1034 static struct user_namespace *netns_owner(struct ns_common *ns) 1035 { 1036 return to_net_ns(ns)->user_ns; 1037 } 1038 1039 const struct proc_ns_operations netns_operations = { 1040 .name = "net", 1041 .type = CLONE_NEWNET, 1042 .get = netns_get, 1043 .put = netns_put, 1044 .install = netns_install, 1045 .owner = netns_owner, 1046 }; 1047 #endif 1048